LLM Reference
activeReasoningResearch

GeneBench-Pro

Metric: Pass rate (higher is better)Introduced: 2026

Research-level computational-biology benchmark for AI agents performing judgment-heavy, multistage scientific analyses in genomics, quantitative biology, and translational biomedicine. The suite contains 129 synthetically constructed problems across 10 primary domains and 21 subdomains, with known causal structure and deterministic grading against decision-relevant target estimands. High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

25

tracked on this benchmark

Score band

31.5 – 0.6

best → lowest tracked

Snapshot trend

+24.08

Jun 30 → Jul 9 · 1 models

Leaderboard

Tracked models ranked by Pass rate (higher is better).

Compare candidates
#Model variant and provenanceScore
1
GPT-5.6 Sol
Version: Pro mode enabled / GPT Pro (Extended); max reasoning; GA launch cites 28.7% standard modeHarness: Not recordedEvaluator: Not recordedObserved: Jul 9, 2026Confidence: Not recordedSource

Notes: Keeps the higher Pro-mode score (31.5) per thinking-model rules; GA launch table also cites 28.7% in standard mode.

31.5
2
GPT-5.6 Terra
Version: GPT Pro (Extended); max reasoning comparatorHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
28.5
3
GPT-5.6 Luna
Version: GPT Pro (Extended); max reasoning comparatorHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
23.6
4
GPT-5.5 Pro
Version: GPT Pro (Extended)Harness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
20.5
5
GPT-5.4 Pro
Version: GPT Pro (Extended)Harness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
16.3
6
Claude Opus 4.8
Version: maxHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
16.0
7
GPT-5.5
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
12.0
8
GPT-5.4
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
8.9
9
GPT-5.2 Pro
Version: GPT Pro (Extended)Harness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
8.5
10
Gemini 3.5 Flash
Version: highHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
8.1
11
GPT-5.2
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
4.9
12
GLM-5.2
Version: highHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
4.6
13
Kimi K2.6
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
4.4
14
Qwen3.7-Max
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
4.0
15
DeepSeek V4 Pro
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
2.4
16
DeepSeek V4 Flash
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
2.4
17
Qwen3.7-Plus
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
2.3
18
Kimi K2.7-Code
Version: reasoning_enabledHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
2.3
19
Xiaomi MiMo-V2.5-Pro
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
2.0
20
Grok 4.3
Version: highHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
1.5
21
GLM-5.1
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
1.2
22
Xiaomi MiMo-V2.5
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
1.2
23
MiniMax M3
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
0.9
24
Hunyuan Hy3 Preview
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
0.9
25
MiniMax M2.7
Version: xhighHarness: Not recordedEvaluator: Not recordedObserved: Jun 30, 2026Confidence: Not recordedSource
0.6

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the GeneBench-Pro benchmark measure?

Research-level computational-biology benchmark for AI agents performing judgment-heavy, multistage scientific analyses in genomics, quantitative biology, and translational biomedicine. The suite contains 129 synthetically constructed problems across 10 primary domains and 21 subdomains, with known causal structure and deterministic grading against decision-relevant target estimands. On this page it lists 25 tracked model variants where higher is better.

Is a higher GeneBench-Pro score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this GeneBench-Pro data?

This benchmark was last reviewed on Jul 2, 2026. The tracked score average moved +24.08 points across the last 2 snapshots.

Related benchmarks

Last reviewed: Jul 2, 2026

Resources