# Vals Index / tax agent bench > Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/vals-ai%3Avals_index%3Atax_agent_bench. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/vals-ai%3Avals_index%3Atax_agent_bench. A single measure of AI's potential economic impact — agentic model performance across finance, coding, legal, and tax tasks, weighted by each sector's share of U.S. GDP. Canonical URL: https://benchize.com/benchmarks/vals-ai%3Avals_index%3Atax_agent_bench 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 single measure of AI's potential economic impact — agentic model performance across finance, coding, legal, and tax tasks, weighted by each sector's share of U.S. GDP. Composite indexes · Models tested: 14 · Score publishers: Vals AI ## Model results | # | Model | Score | ± | Cost | Latency | | | --- | --- | --- | --- | --- | --- | --- | | 1 | [Claude Opus 5.5](https://benchize.com/models/claude-opus-5-5-max.md)Max | 45.57% | SE ±3.59 pp | $15.09 | 4,012.55source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 45.573 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 15.086184 \| USD \| per_test \| Lower is better \| \| `latency` \| 4012.548 \| 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: ±3.59 pp sourcePopulation 23 stderr 3.593 tokenTotals {"cache_read_tokens":648981098,"cache_write_tokens":166280701,"input_tokens":815304119,"output_tokens":88000495,"reasoning_tokens":76703031} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label anthropic/claude-opus-5-5 Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/anthropic~1claude-opus-5-5` | | | | | | | | 2 | [Gemini 4 Argon](https://benchize.com/models/gemini-4-argon-high.md)High | 44.56% | SE ±3.59 pp | $3.47 | 985.63source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 44.56 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 3.473645 \| USD \| per_test \| Lower is better \| \| `latency` \| 985.634 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 262144 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±3.59 pp sourcePopulation 23 stderr 3.587 tokenTotals {"cache_read_tokens":200405700,"cache_write_tokens":null,"input_tokens":285519334,"output_tokens":14493886,"reasoning_tokens":6146392} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label google/gemini-4-argon Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/google~1gemini-4-argon` | | | | | | | | 3 | [DeepSeek V4.1 Flash 1017](https://benchize.com/models/deepseek-v4-1-flash-1017-max.md)Max | 44.35% | SE ±3.57 pp | $0.46 | 1,264.43source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 44.35 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.459454 \| USD \| per_test \| Lower is better \| \| `latency` \| 1264.434 \| 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 1 topP 1 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±3.57 pp sourcePopulation 23 stderr 3.574 tokenTotals {"cache_read_tokens":712066688,"cache_write_tokens":null,"input_tokens":779160448,"output_tokens":53561725,"reasoning_tokens":45716077} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label deepseek/deepseek-v4.1-flash-1017 Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/deepseek~1deepseek-v4.1-flash-1017` | | | | | | | | 4 | [Claude Sonnet 5.5](https://benchize.com/models/claude-sonnet-5-5-max.md)Max | 41.96% | SE ±3.56 pp | $9.25 | 3,560.54source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 41.959 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 9.245466 \| USD \| per_test \| Lower is better \| \| `latency` \| 3560.542 \| 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: ±3.56 pp sourcePopulation 23 stderr 3.561 tokenTotals {"cache_read_tokens":954147139,"cache_write_tokens":196666034,"input_tokens":1150865417,"output_tokens":108029684,"reasoning_tokens":94199819} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label anthropic/claude-sonnet-5-5 Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/anthropic~1claude-sonnet-5-5` | | | | | | | | 5 | [Muse Wonder](https://benchize.com/models/muse-wonder-max.md)Max | 36.27% | SE ±3.47 pp | $1.88 | 492.84source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 36.269 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 1.884439 \| USD \| per_test \| Lower is better \| \| `latency` \| 492.843 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 131072 provider Meta reasoning Not reported Reasoning effort max Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±3.47 pp sourcePopulation 23 stderr 3.47 tokenTotals {"cache_read_tokens":187857044,"cache_write_tokens":0,"input_tokens":225535207,"output_tokens":8770631,"reasoning_tokens":5466021} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label meta/muse-wonder Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/meta~1muse-wonder` | | | | | | | | 6 | [MiMo-V2.6-Pro](https://benchize.com/models/mimo-v2-6-pro.md) | 32.39% | SE ±3.36 pp | $0.13 | 1,547.7source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 32.388 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.134306 \| USD \| per_test \| Lower is better \| \| `latency` \| 1547.695 \| 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: ±3.36 pp sourcePopulation 23 stderr 3.361 tokenTotals {"cache_read_tokens":102894848,"cache_write_tokens":null,"input_tokens":136588210,"output_tokens":12521886,"reasoning_tokens":8284595} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label xiaomi/mimo-v2.6-pro Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/xiaomi~1mimo-v2.6-pro` | | | | | | | | 7 | [MiniMax M3.1 Flash](https://benchize.com/models/minimax-m3-1-flash-max.md)Max | 32.14% | SE ±3.34 pp | $0.32 | 1,962.77source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 32.136 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.320948 \| USD \| per_test \| Lower is better \| \| `latency` \| 1962.771 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 262144 provider MiniMax reasoning Not reported Reasoning effort max Temperature 1 topP 0.95 verbosity Not reported ### Reported statistics Uncertainty Standard error: ±3.34 pp sourcePopulation 23 stderr 3.342 tokenTotals {"cache_read_tokens":466865454,"cache_write_tokens":0,"input_tokens":551883325,"output_tokens":70312257,"reasoning_tokens":62154212} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label minimax/MiniMax-M3.1-flash Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/minimax~1MiniMax-M3.1-flash` | | | | | | | | 8 | [Gemini 3.8 Flash](https://benchize.com/models/gemini-3-8-flash-high.md)High | 32.12% | SE ±3.37 pp | $0.86 | 178.33source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 32.124 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.857214 \| USD \| per_test \| Lower is better \| \| `latency` \| 178.329 \| 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: ±3.37 pp sourcePopulation 23 stderr 3.37 tokenTotals {"cache_read_tokens":208172664,"cache_write_tokens":null,"input_tokens":262586790,"output_tokens":7012701,"reasoning_tokens":4789793} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label google/gemini-3.8-flash Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/google~1gemini-3.8-flash` | | | | | | | | 9 | [Claude Haiku 5.5](https://benchize.com/models/claude-haiku-5-5-max.md)Max | 31.92% | SE ±3.35 pp | $1.16 | 2,973.65source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 31.922 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 1.164207 \| USD \| per_test \| Lower is better \| \| `latency` \| 2973.647 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort max harness Not reported maxOutputTokens 128000 provider Anthropic reasoning Not reported Reasoning effort max Temperature Not reported topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±3.35 pp sourcePopulation 23 stderr 3.352 tokenTotals {"cache_read_tokens":613966554,"cache_write_tokens":168173092,"input_tokens":782193467,"output_tokens":143162719,"reasoning_tokens":133330984} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label anthropic/claude-haiku-5-5 Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/anthropic~1claude-haiku-5-5` | | | | | | | | 10 | [MiMo-V2.6-Flash](https://benchize.com/models/mimo-v2-6-flash.md) | 28.51% | SE ±3.14 pp | $0.05 | 1,038.93source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 28.507 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.048982 \| USD \| per_test \| Lower is better \| \| `latency` \| 1038.928 \| 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: ±3.14 pp sourcePopulation 23 stderr 3.14 tokenTotals {"cache_read_tokens":127944128,"cache_write_tokens":null,"input_tokens":168296091,"output_tokens":12132305,"reasoning_tokens":7352234} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label xiaomi/mimo-v2.6-flash Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/xiaomi~1mimo-v2.6-flash` | | | | | | | | 11 | [DeepSeek V4 Pro 0813](https://benchize.com/models/deepseek-v4-pro-0813-max.md)Max | 26.76% | SE ±3.09 pp | $0.73 | 1,596.76source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 26.756 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.729136 \| USD \| per_test \| Lower is better \| \| `latency` \| 1596.755 \| 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: ±3.09 pp sourcePopulation 23 stderr 3.093 tokenTotals {"cache_read_tokens":133313152,"cache_write_tokens":null,"input_tokens":172298347,"output_tokens":21059828,"reasoning_tokens":17608474} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label deepseek/deepseek-v4-pro-0813 Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/deepseek~1deepseek-v4-pro-0813` | | | | | | | | 12 | [GPT-6.1 Sol](https://benchize.com/models/gpt-6-1-sol-max.md)Max | 21.75% | SE ±2.98 pp | $1.32 | 1,213.88source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 21.754 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 1.323601 \| USD \| per_test \| Lower is better \| \| `latency` \| 1213.878 \| 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: ±2.98 pp sourcePopulation 23 stderr 2.977 tokenTotals {"cache_read_tokens":103439338,"cache_write_tokens":45805894,"input_tokens":149308416,"output_tokens":12717430,"reasoning_tokens":9488118} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label openai/gpt-6.1-sol Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/openai~1gpt-6.1-sol` | | | | | | | | 13 | [GPT-6 Astra](https://benchize.com/models/gpt-6-astra-max.md)Max | 20.72% | SE ±2.92 pp | $5.8 | 900.59source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 20.717 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 5.7963 \| USD \| per_test \| Lower is better \| \| `latency` \| 900.586 \| 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: ±2.92 pp sourcePopulation 23 stderr 2.924 tokenTotals {"cache_read_tokens":71528751,"cache_write_tokens":null,"input_tokens":115812555,"output_tokens":12032734,"reasoning_tokens":9165100} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label openai/gpt-6-astra Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/openai~1gpt-6-astra` | | | | | | | | 14 | [GPT-6 Luna](https://benchize.com/models/gpt-6-luna-max.md)Max | 19.95% | SE ±2.87 pp | $0.19 | 1,995.74source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 19.947 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.190956 \| USD \| per_test \| Lower is better \| \| `latency` \| 1995.736 \| 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: ±2.87 pp sourcePopulation 23 stderr 2.866 tokenTotals {"cache_read_tokens":368052701,"cache_write_tokens":80918209,"input_tokens":449103207,"output_tokens":40664272,"reasoning_tokens":37199155} ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_vals_index.ye6C1wH2.json) Source Vals AI Original model label openai/gpt-6-luna Source version 2.2 Observed 2026-10-21T16:45:49.430Z Evidence SHA-256 `0e95aee94cd2bf34bc4b43a8f4bb16c7817de138613c21011873fea4e0786b45` Source locator `/tasks/tax_agent_bench/openai~1gpt-6-luna` | | | | | | | --- Cite as: Vals Index / tax agent bench — Benchize, https://benchize.com/benchmarks/vals-ai%3Avals_index%3Atax_agent_bench (retrieved 2026-10-21)