# TaxEval v2 / stepwise reasoning > Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/vals-ai%3Atax_eval_v2%3Astepwise_reasoning. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/vals-ai%3Atax_eval_v2%3Astepwise_reasoning. A Vals-created set of questions and responses to tax questions Canonical URL: https://benchize.com/benchmarks/vals-ai%3Atax_eval_v2%3Astepwise_reasoning Last updated: Wed, 21 Oct 2026 18:52:20 GMT Reproduction and quotation must name Benchize, link to the canonical HTML URL, and retain individual source citations. A Vals-created set of questions and responses to tax questions Professional work · Models tested: 40 · Score publishers: Vals AI ## Model results | # | Model | Score | ± | Cost | Latency | | | --- | --- | --- | --- | --- | --- | --- | | 1 | [Muse Spark 1.2](https://benchize.com/models/muse-spark-1-2-xhigh.md)Extra high | 92.72% | SE ±0.74 pp | $0.01 | 48.31source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 92.723 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.014199 \| USD \| per_test \| Lower is better \| \| `latency` \| 48.313 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 131072 provider Meta reasoning Not reported Reasoning effort xhigh Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.74 pp sourcePopulation 145 stderr 0.743 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label meta/muse_spark_1_2 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/meta~1muse_spark_1_2` | | | | | | | | 2 | [Muse Spark 1.1](https://benchize.com/models/muse-spark-1-1-xhigh.md)Extra high | 92.56% | SE ±0.75 pp | $0.01 | 32.82source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 92.559 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.014921 \| USD \| per_test \| Lower is better \| \| `latency` \| 32.816 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 256000 provider Meta reasoning Not reported Reasoning effort xhigh Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.75 pp sourcePopulation 145 stderr 0.75 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label meta/muse_spark_1_1 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/meta~1muse_spark_1_1` | | | | | | | | 3 | [Claude Sonnet 4.6](https://benchize.com/models/claude-sonnet-4-6-max.md)Max | 88.39% | SE ±0.92 pp | $0.13 | 127.88source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 88.389 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.127973 \| USD \| per_test \| Lower is better \| \| `latency` \| 127.879 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning true Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.92 pp sourcePopulation 145 stderr 0.916 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-sonnet-4-6 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-sonnet-4-6` | | | | | | | | 4 | [Claude Opus 5](https://benchize.com/models/claude-opus-5.md) | 88.31% | SE ±0.92 pp | $0.11 | 58.34source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 88.307 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.114274 \| USD \| per_test \| Lower is better \| \| `latency` \| 58.339 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens Not reported provider Anthropic reasoning Not reported Reasoning effort unknown Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.92 pp sourcePopulation 145 stderr 0.919 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-opus-5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-opus-5` | | | | | | | | 5 | [Claude Fable 5](https://benchize.com/models/claude-fable-5-max.md)Max | 88.06% | SE ±0.93 pp | $0.22 | 56.87source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 88.062 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.220133 \| USD \| per_test \| Lower is better \| \| `latency` \| 56.865 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.93 pp sourcePopulation 145 stderr 0.927 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-fable-5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-fable-5` | | | | | | | | 6 | [Claude Fable 5.1](https://benchize.com/models/claude-fable-5-1-max.md)Max | 87.82% | SE ±0.94 pp | $0.35 | 85.56source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 87.817 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.346549 \| USD \| per_test \| Lower is better \| \| `latency` \| 85.564 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.94 pp sourcePopulation 145 stderr 0.935 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-fable-5-1 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-fable-5-1` | | | | | | | | 7 | [Claude Sonnet 5](https://benchize.com/models/claude-sonnet-5-max.md)Max | 87.24% | SE ±0.95 pp | $0.3 | 202.49source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 87.244 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.301619 \| USD \| per_test \| Lower is better \| \| `latency` \| 202.488 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.95 pp sourcePopulation 145 stderr 0.954 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-sonnet-5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-sonnet-5` | | | | | | | | 8 | [Gemini 3.5 Flash](https://benchize.com/models/gemini-3-5-flash-high.md)High | 86.84% | SE ±0.97 pp | $0.04 | 19.69source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.836 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.038136 \| USD \| per_test \| Lower is better \| \| `latency` \| 19.694 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.97 pp sourcePopulation 145 stderr 0.967 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.5-flash Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.5-flash` | | | | | | | | 9 | [Claude Opus 4.7](https://benchize.com/models/claude-opus-4-7-max.md)Max | 86.75% | SE ±0.97 pp | $0.12 | 62.07source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.754 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.124592 \| USD \| per_test \| Lower is better \| \| `latency` \| 62.071 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.97 pp sourcePopulation 145 stderr 0.969 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-opus-4-7 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-opus-4-7` | | | | | | | | 9 | [Claude Opus 4.8](https://benchize.com/models/claude-opus-4-8-max.md)Max | 86.75% | SE ±0.97 pp | $0.16 | 70.33source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.754 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.159139 \| USD \| per_test \| Lower is better \| \| `latency` \| 70.33 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.97 pp sourcePopulation 145 stderr 0.969 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label anthropic/claude-opus-4-8 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/anthropic~1claude-opus-4-8` | | | | | | | | 11 | [GPT-5.6 Terra](https://benchize.com/models/gpt-5-6-terra-xhigh.md)Extra high | 86.51% | SE ±0.98 pp | $0.04 | 37.28source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.509 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.035909 \| USD \| per_test \| Lower is better \| \| `latency` \| 37.28 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider OpenAI reasoning Not reported Reasoning effort xhigh Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.98 pp sourcePopulation 145 stderr 0.977 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-5.6-terra Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-5.6-terra` | | | | | | | | 12 | [Gemini 3.8 Flash](https://benchize.com/models/gemini-3-8-flash-high.md)High | 86.43% | SE ±0.98 pp | $0.02 | 8.76source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.427 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.023535 \| USD \| per_test \| Lower is better \| \| `latency` \| 8.759 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.98 pp sourcePopulation 145 stderr 0.979 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.8-flash Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.8-flash` | | | | | | | | 13 | [Gemini 3.6 Flash](https://benchize.com/models/gemini-3-6-flash-high.md)High | 86.18% | SE ±0.99 pp | $0.02 | 12.34source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 86.182 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.020385 \| USD \| per_test \| Lower is better \| \| `latency` \| 12.335 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.99 pp sourcePopulation 145 stderr 0.987 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.6-flash Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.6-flash` | | | | | | | | 14 | [Gemini 3.7 Flash](https://benchize.com/models/gemini-3-7-flash-high.md)High | 85.85% | SE ±1 pp | $0.02 | 7.83source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 85.854 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.018681 \| USD \| per_test \| Lower is better \| \| `latency` \| 7.832 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1 pp sourcePopulation 145 stderr 0.997 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.7-flash Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.7-flash` | | | | | | | | 15 | [GPT-5.6 Luna](https://benchize.com/models/gpt-5-6-luna-max.md)Max | 85.77% | SE ±1 pp | $0.01 | 76.52source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 85.773 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.012217 \| USD \| per_test \| Lower is better \| \| `latency` \| 76.52 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider OpenAI reasoning Not reported Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1 pp sourcePopulation 145 stderr 0.999 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-5.6-luna Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-5.6-luna` | | | | | | | | 16 | [Gemini 3 Flash Preview](https://benchize.com/models/gemini-3-flash-preview-high.md)High | 85.45% | SE ±1.01 pp | $0.00576 | 12source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 85.446 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.005764 \| USD \| per_test \| Lower is better \| \| `latency` \| 12.001 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.01 pp sourcePopulation 145 stderr 1.008 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3-flash-preview Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3-flash-preview` | | | | | | | | 16 | [MiniMax M3](https://benchize.com/models/minimax-m3.md) | 85.45% | SE ±1.01 pp | $0.00728 | 94.04source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 85.446 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.00728 \| USD \| per_test \| Lower is better \| \| `latency` \| 94.036 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 512000 provider MiniMax reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.01 pp sourcePopulation 145 stderr 1.008 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label minimax/MiniMax-M3 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/minimax~1MiniMax-M3` | | | | | | | | 18 | [Gemini 3.1 Pro Preview](https://benchize.com/models/gemini-3-1-pro-preview-high.md)High | 85.02% | SE ±1.02 pp | $0.03 | 41.93source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 85.016 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.03149 \| USD \| per_test \| Lower is better \| \| `latency` \| 41.931 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65535 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.02 pp sourcePopulation 145 stderr 1.019 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.1-pro-preview Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.1-pro-preview` | | | | | | | | 19 | [Nemotron 3 Ultra](https://benchize.com/models/nemotron-3-ultra-550b-a55b.md) | 84.55% | SE ±1.03 pp | — | 36.49source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.546 \| percent \| reported \| Higher is better \| \| `latency` \| 36.485 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider Nvidia reasoning Not reported Reasoning effort unknown Temperature Not reported topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.03 pp sourcePopulation 145 stderr 1.034 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label nvidia/nemotron-3-ultra-550b-a55b Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/nvidia~1nemotron-3-ultra-550b-a55b` | | | | | | | | 19 | [Mistral Large 3 2512](https://benchize.com/models/mistral-large-2512.md) | 84.55% | SE ±1.03 pp | $0.0016 | 29.53source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.546 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.001601 \| USD \| per_test \| Lower is better \| \| `latency` \| 29.531 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 256000 provider Mistral AI reasoning Not reported Reasoning effort unknown Temperature 0.065 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.03 pp sourcePopulation 145 stderr 1.034 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label mistralai/mistral-large-2512 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/mistralai~1mistral-large-2512` | | | | | | | | 19 | [GPT-5.6 Sol](https://benchize.com/models/gpt-5-6-sol-max.md)Max | 84.55% | SE ±1.03 pp | $0.14 | 62.03source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.546 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.138418 \| USD \| per_test \| Lower is better \| \| `latency` \| 62.034 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider OpenAI reasoning Not reported Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.03 pp sourcePopulation 145 stderr 1.034 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-5.6-sol Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-5.6-sol` | | | | | | | | 22 | [MiMo-V2.5-Pro](https://benchize.com/models/mimo-v2-5-pro.md) | 84.46% | SE ±1.04 pp | $0.00255 | 62.83source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.464 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.002547 \| USD \| per_test \| Lower is better \| \| `latency` \| 62.829 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider Xiaomi reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.04 pp sourcePopulation 145 stderr 1.036 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label xiaomi/mimo-v2.5-pro Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/xiaomi~1mimo-v2.5-pro` | | | | | | | | 22 | [GPT-5.5](https://benchize.com/models/gpt-5-5-xhigh.md)Extra high | 84.46% | SE ±1.04 pp | $0.06 | 95.83source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.464 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.06042 \| USD \| per_test \| Lower is better \| \| `latency` \| 95.83 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider OpenAI reasoning Not reported Reasoning effort xhigh Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.04 pp sourcePopulation 145 stderr 1.036 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-5.5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-5.5` | | | | | | | | 24 | [GPT-4o (2024-11-20)](https://benchize.com/models/gpt-4o-2024-11-20-high.md)High | 84.22% | SE ±1.04 pp | $0.00581 | 5.86source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 84.219 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.005806 \| USD \| per_test \| Lower is better \| \| `latency` \| 5.855 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 16384 provider OpenAI reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.04 pp sourcePopulation 145 stderr 1.042 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-4o-2024-11-20 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-4o-2024-11-20` | | | | | | | | 25 | [DeepSeek V4 Pro](https://benchize.com/models/deepseek-v4-pro-max.md)Max | 83.57% | SE ±1.06 pp | $0.03 | 189.7source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 83.565 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.033187 \| USD \| per_test \| Lower is better \| \| `latency` \| 189.695 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider DeepSeek reasoning Not reported Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.06 pp sourcePopulation 145 stderr 1.06 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label deepseek/deepseek-v4-pro Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/deepseek~1deepseek-v4-pro` | | | | | | | | 26 | [Gemini 3.5 Flash-Lite](https://benchize.com/models/gemini-3-5-flash-lite-high.md)High | 83.48% | SE ±1.06 pp | $0.00729 | 8.82source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 83.483 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.00729 \| USD \| per_test \| Lower is better \| \| `latency` \| 8.815 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.06 pp sourcePopulation 145 stderr 1.062 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.5-flash-lite Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.5-flash-lite` | | | | | | | | 26 | [DeepSeek V4 Pro 0813](https://benchize.com/models/deepseek-v4-pro-0813-max.md)Max | 83.48% | SE ±1.06 pp | $0.05 | 226.3source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 83.483 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.045816 \| USD \| per_test \| Lower is better \| \| `latency` \| 226.296 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 384000 provider DeepSeek reasoning Not reported Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.06 pp sourcePopulation 145 stderr 1.062 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label deepseek/deepseek-v4-pro-0813 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/deepseek~1deepseek-v4-pro-0813` | | | | | | | | 28 | [Gemini 3.1 Flash Lite Preview](https://benchize.com/models/gemini-3-1-flash-lite-preview-high.md)High | 83.16% | SE ±1.07 pp | $0.00096 | 7.64source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 83.156 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.00096 \| USD \| per_test \| Lower is better \| \| `latency` \| 7.644 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.07 pp sourcePopulation 145 stderr 1.07 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-3.1-flash-lite-preview Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-3.1-flash-lite-preview` | | | | | | | | 29 | [MiMo-V2.5](https://benchize.com/models/mimo-v2-5.md) | 82.91% | SE ±1.08 pp | $0.000729 | 26.3source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 82.911 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000729 \| USD \| per_test \| Lower is better \| \| `latency` \| 26.3 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider Xiaomi reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.08 pp sourcePopulation 145 stderr 1.076 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label xiaomi/mimo-v2.5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/xiaomi~1mimo-v2.5` | | | | | | | | 30 | [GPT-4o (2024-08-06)](https://benchize.com/models/gpt-4o-2024-08-06-high.md)High | 81.44% | SE ±1.11 pp | $0.00442 | 9.54source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 81.439 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.004417 \| USD \| per_test \| Lower is better \| \| `latency` \| 9.543 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 16384 provider OpenAI reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.11 pp sourcePopulation 145 stderr 1.112 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-4o-2024-08-06 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-4o-2024-08-06` | | | | | | | | 31 | [DeepSeek V4 Flash 0731](https://benchize.com/models/deepseek-v4-flash-0731-high.md)High | 80.87% | SE ±1.13 pp | $0.00976 | 72.2source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 80.867 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.009755 \| USD \| per_test \| Lower is better \| \| `latency` \| 72.203 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 384000 provider DeepSeek reasoning Not reported Reasoning effort high Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.13 pp sourcePopulation 145 stderr 1.125 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label deepseek/deepseek-v4-flash-0731 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/deepseek~1deepseek-v4-flash-0731` | | | | | | | | 31 | [MiniMax M2.5](https://benchize.com/models/minimax-m2-5.md) | 80.87% | SE ±1.13 pp | $0.00863 | 73.24source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 80.867 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.008629 \| USD \| per_test \| Lower is better \| \| `latency` \| 73.243 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 196608 provider MiniMax reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.13 pp sourcePopulation 145 stderr 1.125 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label minimax/MiniMax-M2.5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/minimax~1MiniMax-M2.5` | | | | | | | | 33 | [Mistral Medium 3.5](https://benchize.com/models/mistral-medium-3-5-high.md)High | 79.23% | SE ±1.16 pp | $0.08 | 102.39source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 79.231 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.079101 \| USD \| per_test \| Lower is better \| \| `latency` \| 102.392 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 80000 provider Mistral AI reasoning Not reported Reasoning effort high Temperature 0.7 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.16 pp sourcePopulation 145 stderr 1.16 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label mistralai/mistral-medium-3.5 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/mistralai~1mistral-medium-3.5` | | | | | | | | 34 | [MiniMax M2.7](https://benchize.com/models/minimax-m2-7.md) | 78.99% | SE ±1.17 pp | $0.00347 | 28.69source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 78.986 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.003474 \| USD \| per_test \| Lower is better \| \| `latency` \| 28.694 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 196608 provider MiniMax reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.17 pp sourcePopulation 145 stderr 1.165 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label minimax/MiniMax-M2.7 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/minimax~1MiniMax-M2.7` | | | | | | | | 35 | [Gemini 2.5 Flash Lite Preview 09-2025](https://benchize.com/models/gemini-2-5-flash-lite-preview-09-2025.md) | 78.5% | SE ±1.18 pp | $0.000477 | 3.6source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 78.496 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000477 \| USD \| per_test \| Lower is better \| \| `latency` \| 3.597 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65535 provider Google reasoning Not reported Reasoning effort unknown Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.18 pp sourcePopulation 145 stderr 1.175 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label google/gemini-2.5-flash-lite-preview-09-2025 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/google~1gemini-2.5-flash-lite-preview-09-2025` | | | | | | | | 36 | [MiniMax M2.1](https://benchize.com/models/minimax-m2-1.md) | 78.25% | SE ±1.18 pp | $0.00914 | 113.26source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 78.25 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.009137 \| USD \| per_test \| Lower is better \| \| `latency` \| 113.258 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 131000 provider MiniMax reasoning Not reported Reasoning effort unknown Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.18 pp sourcePopulation 145 stderr 1.18 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label minimax/MiniMax-M2.1 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/minimax~1MiniMax-M2.1` | | | | | | | | 37 | [Command A](https://benchize.com/models/command-a.md) | 73.75% | SE ±1.26 pp | $0.00648 | 11.21source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 73.753 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.006484 \| USD \| per_test \| Lower is better \| \| `latency` \| 11.214 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 8000 provider Cohere reasoning Not reported Reasoning effort unknown Temperature 0.3 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.26 pp sourcePopulation 145 stderr 1.258 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label cohere/command-a-03-2025 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/cohere~1command-a-03-2025` | | | | | | | | 38 | [GPT-4o-mini (2024-07-18)](https://benchize.com/models/gpt-4o-mini-2024-07-18-high.md)High | 71.55% | SE ±1.29 pp | $0.000323 | 8.9source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 71.545 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000323 \| USD \| per_test \| Lower is better \| \| `latency` \| 8.899 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 16384 provider OpenAI reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.29 pp sourcePopulation 145 stderr 1.29 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label openai/gpt-4o-mini-2024-07-18 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/openai~1gpt-4o-mini-2024-07-18` | | | | | | | | 39 | [Laguna Xs 2](https://benchize.com/models/laguna-xs-2.md) | 71.22% | SE ±1.3 pp | $0.000305 | 27.45source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 71.218 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000305 \| USD \| per_test \| Lower is better \| \| `latency` \| 27.449 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens Not reported provider Poolside reasoning Not reported Reasoning effort unknown Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.3 pp sourcePopulation 145 stderr 1.295 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label poolside/laguna-xs.2 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/poolside~1laguna-xs.2` | | | | | | | | 40 | [Laguna M 1](https://benchize.com/models/laguna-m-1.md) | 1.8% | SE ±0.38 pp | — | 212.64source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 1.799 \| percent \| reported \| Higher is better \| \| `latency` \| 212.643 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 32768 provider Poolside reasoning Not reported Reasoning effort unknown Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.38 pp sourcePopulation 145 stderr 0.38 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_tax_eval_v2.B3yoffr8.json) Source Vals AI Original model label poolside/laguna-m.1 Source version 2 Observed 2026-10-21T16:45:59.000Z Evidence SHA-256 `8b354a6ae848f5bb25a2e8b061b5f576d5b9e0c11b941dc1e6bed078e6529cd6` Source locator `/tasks/stepwise_reasoning/poolside~1laguna-m.1` | | | | | | | --- Cite as: TaxEval v2 / stepwise reasoning — Benchize, https://benchize.com/benchmarks/vals-ai%3Atax_eval_v2%3Astepwise_reasoning (retrieved 2026-10-21)