# MedQA / black > Source: Benchize (https://benchize.com) — https://benchize.com/benchmarks/vals-ai%3Amedqa%3Ablack. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/benchmarks/vals-ai%3Amedqa%3Ablack. Evaluating language model bias in medical questions. Canonical URL: https://benchize.com/benchmarks/vals-ai%3Amedqa%3Ablack 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 language model bias in medical questions. Professional work · Models tested: 12 · Score publishers: Vals AI ## Model results | # | Model | Score | ± | Cost | Latency | | | --- | --- | --- | --- | --- | --- | --- | | 1 | [Gemini 3.1 Pro Preview](https://benchize.com/models/gemini-3-1-pro-preview-high.md)High | 96.8% | SE ±0.39 pp | $0.02 | 39.84source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 96.8 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.018058 \| USD \| per_test \| Lower is better \| \| `latency` \| 39.836 \| 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: ±0.39 pp sourcePopulation 95 stderr 0.394 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label google/gemini-3.1-pro-preview Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/google~1gemini-3.1-pro-preview` | | | | | | | | 2 | [Gemini 3 Flash Preview](https://benchize.com/models/gemini-3-flash-preview-high.md)High | 95.7% | SE ±0.45 pp | $0.01 | 27.73source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 95.7 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.013337 \| USD \| per_test \| Lower is better \| \| `latency` \| 27.728 \| 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: ±0.45 pp sourcePopulation 95 stderr 0.454 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label google/gemini-3-flash-preview Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/google~1gemini-3-flash-preview` | | | | | | | | 3 | [Claude Sonnet 4.6](https://benchize.com/models/claude-sonnet-4-6-thinking.md)Thinking | 92.75% | SE ±0.58 pp | $0.04 | 36.47source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 92.75 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.036805 \| USD \| per_test \| Lower is better \| \| `latency` \| 36.466 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 128000 provider Anthropic reasoning true Reasoning effort unknown Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±0.58 pp sourcePopulation 95 stderr 0.58 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label anthropic/claude-sonnet-4-6 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/anthropic~1claude-sonnet-4-6` | | | | | | | | 4 | [MiniMax M2.5](https://benchize.com/models/minimax-m2-5.md) | 92.1% | SE ±0.6 pp | $0.00302 | 34.79source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 92.1 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.003019 \| USD \| per_test \| Lower is better \| \| `latency` \| 34.785 \| 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: ±0.6 pp sourcePopulation 95 stderr 0.603 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label minimax/MiniMax-M2.5 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/minimax~1MiniMax-M2.5` | | | | | | | | 5 | [MiniMax M2.1](https://benchize.com/models/minimax-m2-1.md) | 91.15% | SE ±0.64 pp | $0.00238 | 36.27source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 91.15 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.00238 \| USD \| per_test \| Lower is better \| \| `latency` \| 36.27 \| 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: ±0.64 pp sourcePopulation 95 stderr 0.635 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label minimax/MiniMax-M2.1 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/minimax~1MiniMax-M2.1` | | | | | | | | 6 | [GPT-4 Turbo](https://benchize.com/models/gpt-4-turbo-high.md)High | 87.81% | SE ±1.44 pp | $0.00955 | 3.14source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 87.813 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.009554 \| USD \| per_test \| Lower is better \| \| `latency` \| 3.14 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 4096 provider OpenAI reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±1.44 pp sourcePopulation 95 stderr 1.436 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label openai/gpt-4-turbo Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/openai~1gpt-4-turbo` | | | | | | | | 6 | [GPT-4o (2024-08-06)](https://benchize.com/models/gpt-4o-2024-08-06-high.md)High | 87.81% | SE ±1.44 pp | $0.00281 | 3.14source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 87.813 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.002807 \| USD \| per_test \| Lower is better \| \| `latency` \| 3.14 \| 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: ±1.44 pp sourcePopulation 95 stderr 1.436 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label openai/gpt-4o-2024-08-06 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/openai~1gpt-4o-2024-08-06` | | | | | | | | 8 | [Mistral Large 3 2512](https://benchize.com/models/mistral-large-2512.md) | 81.85% | SE ±0.86 pp | $0.000783 | 17.94source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 81.85 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000783 \| USD \| per_test \| Lower is better \| \| `latency` \| 17.944 \| 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: ±0.86 pp sourcePopulation 95 stderr 0.862 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label mistralai/mistral-large-2512 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/mistralai~1mistral-large-2512` | | | | | | | | 9 | [Gemini 2.5 Flash Lite Preview 09-2025](https://benchize.com/models/gemini-2-5-flash-lite-preview-09-2025.md) | 81.15% | SE ±0.88 pp | $0.000283 | 5.95source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 81.15 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000283 \| USD \| per_test \| Lower is better \| \| `latency` \| 5.95 \| 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: ±0.88 pp sourcePopulation 95 stderr 0.875 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label google/gemini-2.5-flash-lite-preview-09-2025 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/google~1gemini-2.5-flash-lite-preview-09-2025` | | | | | | | | 10 | [Command A](https://benchize.com/models/command-a.md) | 79.8% | SE ±0.9 pp | $0.00349 | 2.92source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 79.8 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.003489 \| USD \| per_test \| Lower is better \| \| `latency` \| 2.919 \| 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: ±0.9 pp sourcePopulation 95 stderr 0.898 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label cohere/command-a-03-2025 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/cohere~1command-a-03-2025` | | | | | | | | 11 | [GPT-4o-mini (2024-07-18)](https://benchize.com/models/gpt-4o-mini-2024-07-18-high.md)High | 72.47% | SE ±1.96 pp | $0.000157 | 2.27source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 72.467 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000157 \| USD \| per_test \| Lower is better \| \| `latency` \| 2.267 \| 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: ±1.96 pp sourcePopulation 95 stderr 1.959 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label openai/gpt-4o-mini-2024-07-18 Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/openai~1gpt-4o-mini-2024-07-18` | | | | | | | | 12 | [GPT-3.5 Turbo](https://benchize.com/models/gpt-3-5-turbo-high.md)High | 58.73% | SE ±2.16 pp | $0.000428 | 1.77source duration | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 58.726 \| percent \| reported \| Higher is better \| \| `cost_per_test` \| 0.000428 \| USD \| per_test \| Lower is better \| \| `latency` \| 1.771 \| source_duration \| reported_duration \| Lower is better \| ### Configuration computeEffort Not reported harness Not reported maxOutputTokens 4096 provider OpenAI reasoning Not reported Reasoning effort high Temperature 1 topP Not reported verbosity Not reported ### Reported statistics Uncertainty Standard error: ±2.16 pp sourcePopulation 95 stderr 2.159 tokenTotals Not reported ### Source [Original source](https://www.vals.ai/_astro/benchmark_view_medqa.Cco1kvog.json) Source Vals AI Original model label openai/gpt-3.5-turbo Source version 1 Observed 2026-10-21T16:46:00.691Z Evidence SHA-256 `d67905978d932f64698ebe9e230d9519e5bae1c9d7176dda8379dc3874b6f90f` Source locator `/tasks/black/openai~1gpt-3.5-turbo` | | | | | | | --- Cite as: MedQA / black — Benchize, https://benchize.com/benchmarks/vals-ai%3Amedqa%3Ablack (retrieved 2026-10-21)