# MortgageTax / numerical extraction > Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/vals-ai%3Amortgage_tax%3Anumerical_extraction. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/vals-ai%3Amortgage_tax%3Anumerical_extraction. Evaluating reading and understanding tax certificates as images Canonical URL: https://benchize.com/benchmarks/vals-ai%3Amortgage_tax%3Anumerical_extraction 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. Evaluating reading and understanding tax certificates as images Professional work · Models tested: 31 · Score publishers: Vals AI ## Model results | # | Model | Score | ± | Cost | Latency | | | --- | --- | --- | --- | --- | --- | --- | | 1 | [Claude Opus 5](https://benchize.com/models/claude-opus-5.md) | 81.4% | SE ±1.1 pp | $0.07 | 8.48source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 81.399 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.074778 \| USD \| per_test \| Lower is better \| \| `latency` \| 8.479 \| 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: ±1.1 pp sourcePopulation 98 stderr 1.097 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-opus-5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-opus-5` | | | | | | | | 2 | [Claude Opus 4.8](https://benchize.com/models/claude-opus-4-8-max.md)Max | 79.73% | SE ±1.13 pp | $0.13 | 29.84source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 79.73 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.127846 \| USD \| per_test \| Lower is better \| \| `latency` \| 29.835 \| 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: ±1.13 pp sourcePopulation 98 stderr 1.133 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-opus-4-8 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-opus-4-8` | | | | | | | | 3 | [Claude Opus 4.7](https://benchize.com/models/claude-opus-4-7-max.md)Max | 78.7% | SE ±1.15 pp | $0.09 | 19.79source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 78.696 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.093695 \| USD \| per_test \| Lower is better \| \| `latency` \| 19.793 \| 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: ±1.15 pp sourcePopulation 98 stderr 1.154 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-opus-4-7 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-opus-4-7` | | | | | | | | 3 | [Claude Sonnet 5](https://benchize.com/models/claude-sonnet-5-max.md)Max | 78.7% | SE ±1.15 pp | $0.11 | 48source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 78.696 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.110772 \| USD \| per_test \| Lower is better \| \| `latency` \| 47.997 \| 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: ±1.15 pp sourcePopulation 98 stderr 1.154 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-sonnet-5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-sonnet-5` | | | | | | | | 5 | [Gemini 2.5 Pro](https://benchize.com/models/gemini-2-5-pro.md) | 77.58% | SE ±1.18 pp | $0.00465 | 28.25source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 77.583 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.004647 \| USD \| per_test \| Lower is better \| \| `latency` \| 28.249 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 65536 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 98 stderr 1.176 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-2.5-pro Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-2.5-pro` | | | | | | | | 5 | [Claude Fable 5.1](https://benchize.com/models/claude-fable-5-1-max.md)Max | 77.58% | SE ±1.18 pp | $0.18 | 17.78source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 77.583 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.182513 \| USD \| per_test \| Lower is better \| \| `latency` \| 17.784 \| 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: ±1.18 pp sourcePopulation 98 stderr 1.176 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-fable-5-1 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-fable-5-1` | | | | | | | | 7 | [Gemini 3.5 Flash](https://benchize.com/models/gemini-3-5-flash-high.md)High | 77.5% | SE ±1.18 pp | $0.02 | 12.06source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 77.504 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.019197 \| USD \| per_test \| Lower is better \| \| `latency` \| 12.063 \| 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: ±1.18 pp sourcePopulation 98 stderr 1.177 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.5-flash Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.5-flash` | | | | | | | | 8 | [MiniMax M3](https://benchize.com/models/minimax-m3.md) | 77.27% | SE ±1.18 pp | $0.00962 | 39.01source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 77.266 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.009618 \| USD \| per_test \| Lower is better \| \| `latency` \| 39.014 \| 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: ±1.18 pp sourcePopulation 98 stderr 1.182 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label minimax/MiniMax-M3 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/minimax~1MiniMax-M3` | | | | | | | | 9 | [Gemini 3.1 Flash Lite Preview](https://benchize.com/models/gemini-3-1-flash-lite-preview-high.md)High | 77.03% | SE ±1.19 pp | $0.00323 | 6.32source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 77.027 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.003233 \| USD \| per_test \| Lower is better \| \| `latency` \| 6.323 \| 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: ±1.19 pp sourcePopulation 98 stderr 1.186 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.1-flash-lite-preview Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.1-flash-lite-preview` | | | | | | | | 10 | [Gemini 3.6 Flash](https://benchize.com/models/gemini-3-6-flash-high.md)High | 76.79% | SE ±1.19 pp | $0.02 | 11.23source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 76.789 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.023587 \| USD \| per_test \| Lower is better \| \| `latency` \| 11.233 \| 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: ±1.19 pp sourcePopulation 98 stderr 1.19 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.6-flash Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.6-flash` | | | | | | | | 11 | [Claude Fable 5](https://benchize.com/models/claude-fable-5-max.md)Max | 76.71% | SE ±1.19 pp | $0.18 | 17.2source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 76.709 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.183368 \| USD \| per_test \| Lower is better \| \| `latency` \| 17.2 \| 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: ±1.19 pp sourcePopulation 98 stderr 1.192 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-fable-5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-fable-5` | | | | | | | | 12 | [Gemini 3.1 Pro Preview](https://benchize.com/models/gemini-3-1-pro-preview-high.md)High | 76.39% | SE ±1.2 pp | $0.02 | 24.79source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 76.391 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.024731 \| USD \| per_test \| Lower is better \| \| `latency` \| 24.789 \| 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: ±1.2 pp sourcePopulation 98 stderr 1.197 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.1-pro-preview Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.1-pro-preview` | | | | | | | | 13 | [GPT-5.5](https://benchize.com/models/gpt-5-5-xhigh.md)Extra high | 76.15% | SE ±1.2 pp | $0.08 | 39.33source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 76.153 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.079986 \| USD \| per_test \| Lower is better \| \| `latency` \| 39.334 \| 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: ±1.2 pp sourcePopulation 98 stderr 1.201 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-5.5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-5.5` | | | | | | | | 14 | [Gemini 3.5 Flash-Lite](https://benchize.com/models/gemini-3-5-flash-lite-high.md)High | 75.84% | SE ±1.21 pp | $0.00537 | 5.86source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 75.835 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.00537 \| USD \| per_test \| Lower is better \| \| `latency` \| 5.856 \| 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: ±1.21 pp sourcePopulation 98 stderr 1.207 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.5-flash-lite Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.5-flash-lite` | | | | | | | | 15 | [Gemini 3 Flash Preview](https://benchize.com/models/gemini-3-flash-preview-high.md)High | 75.76% | SE ±1.21 pp | $0.00989 | 17.54source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 75.755 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.009888 \| USD \| per_test \| Lower is better \| \| `latency` \| 17.537 \| 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: ±1.21 pp sourcePopulation 98 stderr 1.208 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3-flash-preview Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3-flash-preview` | | | | | | | | 16 | [Gemini 3.7 Flash](https://benchize.com/models/gemini-3-7-flash-high.md)High | 75.68% | SE ±1.21 pp | $0.01 | 4.19source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 75.676 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.01158 \| USD \| per_test \| Lower is better \| \| `latency` \| 4.188 \| 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: ±1.21 pp sourcePopulation 98 stderr 1.21 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.7-flash Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.7-flash` | | | | | | | | 17 | [GPT-5.6 Luna](https://benchize.com/models/gpt-5-6-luna-max.md)Max | 75.6% | SE ±1.21 pp | $0.00614 | 33.26source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 75.596 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.006139 \| USD \| per_test \| Lower is better \| \| `latency` \| 33.26 \| 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: ±1.21 pp sourcePopulation 98 stderr 1.211 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-5.6-luna Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-5.6-luna` | | | | | | | | 18 | [GPT-5.6 Sol](https://benchize.com/models/gpt-5-6-sol-max.md)Max | 75.12% | SE ±1.22 pp | $0.08 | 15.56source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 75.119 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.079634 \| USD \| per_test \| Lower is better \| \| `latency` \| 15.561 \| 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: ±1.22 pp sourcePopulation 98 stderr 1.219 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-5.6-sol Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-5.6-sol` | | | | | | | | 19 | [Claude Sonnet 4.6](https://benchize.com/models/claude-sonnet-4-6-max.md)Max | 74.96% | SE ±1.22 pp | $0.06 | 45.71source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 74.96 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.062771 \| USD \| per_test \| Lower is better \| \| `latency` \| 45.711 \| 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: ±1.22 pp sourcePopulation 98 stderr 1.221 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label anthropic/claude-sonnet-4-6 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/anthropic~1claude-sonnet-4-6` | | | | | | | | 20 | [Muse Spark 1.1](https://benchize.com/models/muse-spark-1-1-xhigh.md)Extra high | 74.48% | SE ±1.23 pp | $0.03 | 30.7source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 74.483 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.025379 \| USD \| per_test \| Lower is better \| \| `latency` \| 30.7 \| 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: ±1.23 pp sourcePopulation 98 stderr 1.229 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label meta/muse_spark_1_1 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/meta~1muse_spark_1_1` | | | | | | | | 21 | [GPT-5.6 Terra](https://benchize.com/models/gpt-5-6-terra-xhigh.md)Extra high | 74.32% | SE ±1.23 pp | $0.02 | 7.02source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 74.324 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.023396 \| USD \| per_test \| Lower is better \| \| `latency` \| 7.021 \| 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: ±1.23 pp sourcePopulation 98 stderr 1.232 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-5.6-terra Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-5.6-terra` | | | | | | | | 22 | [Muse Spark 1.2](https://benchize.com/models/muse-spark-1-2-xhigh.md)Extra high | 73.61% | SE ±1.24 pp | $0.03 | 55.04source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 73.609 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.031564 \| USD \| per_test \| Lower is better \| \| `latency` \| 55.043 \| 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: ±1.24 pp sourcePopulation 98 stderr 1.243 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label meta/muse_spark_1_2 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/meta~1muse_spark_1_2` | | | | | | | | 23 | [GPT-4o (2024-08-06)](https://benchize.com/models/gpt-4o-2024-08-06-high.md)High | 71.15% | SE ±2.5 pp | $0.00678 | 7.27source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 71.145 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.006779 \| USD \| per_test \| Lower is better \| \| `latency` \| 7.274 \| 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: ±2.5 pp sourcePopulation 98 stderr 2.501 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-4o-2024-08-06 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-4o-2024-08-06` | | | | | | | | 24 | [Gemini 3.8 Flash](https://benchize.com/models/gemini-3-8-flash-high.md)High | 70.59% | SE ±1.29 pp | $0.07 | 21.32source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 70.588 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.068848 \| USD \| per_test \| Lower is better \| \| `latency` \| 21.324 \| 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: ±1.29 pp sourcePopulation 98 stderr 1.285 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-3.8-flash Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-3.8-flash` | | | | | | | | 25 | [Gemma 4 31B](https://benchize.com/models/gemma-4-31b-it-high.md)High | 65.66% | SE ±1.34 pp | $0 | 55.5source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 65.66 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0 \| USD \| per_test \| Lower is better \| \| `latency` \| 55.497 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 32768 provider Google reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.34 pp sourcePopulation 98 stderr 1.339 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemma-4-31b-it Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemma-4-31b-it` | | | | | | | | 26 | [GPT-4o (2024-11-20)](https://benchize.com/models/gpt-4o-2024-11-20-high.md)High | 64.87% | SE ±2.63 pp | $0.00694 | 5.85source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 64.865 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.006935 \| USD \| per_test \| Lower is better \| \| `latency` \| 5.85 \| 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: ±2.63 pp sourcePopulation 98 stderr 2.634 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-4o-2024-11-20 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-4o-2024-11-20` | | | | | | | | 27 | [MiMo-V2.5](https://benchize.com/models/mimo-v2-5.md) | 61.05% | SE ±1.38 pp | $0.00162 | 38.22source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 61.049 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.001623 \| USD \| per_test \| Lower is better \| \| `latency` \| 38.218 \| 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: ±1.38 pp sourcePopulation 98 stderr 1.375 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label xiaomi/mimo-v2.5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/xiaomi~1mimo-v2.5` | | | | | | | | 28 | [GPT-4o-mini (2024-07-18)](https://benchize.com/models/gpt-4o-mini-2024-07-18-high.md)High | 60.18% | SE ±2.7 pp | $0.000727 | 5.93source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 60.175 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000727 \| USD \| per_test \| Lower is better \| \| `latency` \| 5.931 \| 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: ±2.7 pp sourcePopulation 98 stderr 2.701 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label openai/gpt-4o-mini-2024-07-18 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/openai~1gpt-4o-mini-2024-07-18` | | | | | | | | 29 | [Gemini 2.5 Flash Lite Preview 09-2025](https://benchize.com/models/gemini-2-5-flash-lite-preview-09-2025.md) | 55.17% | SE ±1.4 pp | $0.000622 | 13.64source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 55.167 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000622 \| USD \| per_test \| Lower is better \| \| `latency` \| 13.637 \| 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: ±1.4 pp sourcePopulation 98 stderr 1.402 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label google/gemini-2.5-flash-lite-preview-09-2025 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/google~1gemini-2.5-flash-lite-preview-09-2025` | | | | | | | | 30 | [Mistral Large 3 2512](https://benchize.com/models/mistral-large-2512.md) | 50% | SE ±1.41 pp | $0.00326 | 9.87source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 50 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.003256 \| USD \| per_test \| Lower is better \| \| `latency` \| 9.865 \| 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: ±1.41 pp sourcePopulation 98 stderr 1.41 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label mistralai/mistral-large-2512 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/mistralai~1mistral-large-2512` | | | | | | | | 31 | [Mistral Medium 3.5](https://benchize.com/models/mistral-medium-3-5-high.md)High | 11.92% | SE ±0.91 pp | $0.03 | 75.47source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 11.924 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.030861 \| USD \| per_test \| Lower is better \| \| `latency` \| 75.468 \| 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: ±0.91 pp sourcePopulation 98 stderr 0.914 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_mortgage_tax.DCd-oO5Y.json) Source Vals AI Original model label mistralai/mistral-medium-3.5 Source version 1 Observed 2026-10-21T16:45:58.500Z Evidence SHA-256 `e5c844561f97d119d66d70269e726ecfb68be79eae39acf14ead3e95422e648c` Source locator `/tasks/numerical_extraction/mistralai~1mistral-medium-3.5` | | | | | | | --- Cite as: MortgageTax / numerical extraction — Benchize, https://benchize.com/benchmarks/vals-ai%3Amortgage_tax%3Anumerical_extraction (retrieved 2026-10-21)