SLR-Bench
Models write Prolog rules that classify trains by their composition. The verifier runs each rule to determine whether its predictions are correct.
- Made by
- Lukas Helff and colleagues
- Released
- June 2025
- Kind
- Abstract puzzles and logic
- Level
- University
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
Gemini 3 Flash #60 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
GPT-5 leads with 81.8% among models released by April 2026 GPT-5 leads with 81.8% among models released by April 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 81.8 GPT-5
- 80.8
Gemini 3 Flash
- 79.5 GPT-5 mini
- 67.5 gpt-oss-120b
- 66.0 GPT-5.4 Mini
- 65.5 GLM-4.7 FP8
- 65.2
Gemini 3.1 Flash-Lite
- 65.0
Gemini 2.5 Flash
- 62.2 o3-mini
- 61.5 o4-mini
- 60.0
Qwen3.5-27B
- 59.8 GPT-5 Chat
- 58.0 Gemini-2.5-pro (snapshot unspecified)
- 57.8
Qwen3.5-35B-A3B
- 55.0 GLM-4.5-Air
- 54.8
Gemma 4 31B IT
- 53.5 Kimi K2.5
- 53.0 Nemotron-3 120B
- 52.8 Kimi K2
- 51.2 gpt-oss-20b
- 51.0
Llama-3.1-Nemotron-Ultra-253B-v1
- 51.0 GPT-5 nano
- 49.2
Nemotron-Cascade 2 30B-A3B
- 49.0
Qwen3.5-9B
- 47.8
Gemini 2.5 Flash-Lite
- 47.0
Ministral 3 14B
- 45.0
Olmo 3 32B Think
- 44.5
Qwen3.5-4B
- 43.8
DeepSeek-R1-Distill-Llama-70B
- 43.8
DeepSeek-R1-Distill-Qwen-32B
- 43.8
Gemma 4 26B A4B
- 41.2 GPT-4.5
- 41.2 Ministral-3 8B
- 40.8
Gemma 4 E4B
- 38.5
Olmo 3.1 32B Think
- 38.2 DeepSeek-R1-Distill Qwen 8B
- 33.8 Ministral-3 3B
- 33.0
Gemma 4 E2B
- 32.2 GPT-4o (snapshot unspecified)
- 28.8 GPT-4 Turbo (snapshot unspecified)
- 27.5 GPT-4o Mini (snapshot unspecified)
- 25.0 Llama-3.3-Nemotron 49B
- 11.5 OLMo-3 7B Think
- 8.20 Kimi-VL A3B
- 3.00 Nemotron-3 30B
- 1 GPT-5 medium 81.8%
- 2
Gemini 3 Flash high 80.8%
- 3 GPT-5 mini high 79.5%
- 4 gpt-oss-120b high 67.5%
- 5 GPT-5.4 Mini medium 66%
- 6 GLM-4.7 FP8 65.5%
- 7
Gemini 3.1 Flash-Lite high 65.2%
- 8
Gemini 2.5 Flash Sep 2025 65%
- 9 o3-mini medium 62.2%
- 10 o4-mini medium 61.5%
- 11
Qwen3.5-27B 60%
- 12 GPT-5 Chat 59.8%
- 13 Gemini-2.5-pro (snapshot unspecified) 58%
- 14
Qwen3.5-35B-A3B 57.8%
- 15 GLM-4.5-Air 55%
- 16
Gemma 4 31B IT 54.8%
- 17 Kimi K2.5 53.5%
- 18 Nemotron-3 120B 53%
- 19 Kimi K2 Sep 2025 52.8%
- 20 gpt-oss-20b high 51.2%
- 21
Llama-3.1-Nemotron-Ultra-253B-v1 51%
- 22 GPT-5 nano high 51%
- 23
Nemotron-Cascade 2 30B-A3B 49.2%
- 24
Qwen3.5-9B 49%
- 25
Gemini 2.5 Flash-Lite Sep 2025, high 47.8%
- 26
Ministral 3 14B 47%
- 27
Olmo 3 32B Think 45%
- 28
Qwen3.5-4B 44.5%
- 29
DeepSeek-R1-Distill-Llama-70B 43.8%
- 30
DeepSeek-R1-Distill-Qwen-32B 43.8%
- 31
Gemma 4 26B A4B 43.8%
- 32 GPT-4.5 41.2%
- 33 Ministral-3 8B 41.2%
- 34
Gemma 4 E4B 40.8%
- 35
Olmo 3.1 32B Think 38.5%
- 36 DeepSeek-R1-Distill Qwen 8B 38.2%
- 37 Ministral-3 3B 33.8%
- 38
Gemma 4 E2B 33%
- 39 GPT-4o (snapshot unspecified) 32.2%
- 40 GPT-4 Turbo (snapshot unspecified) 28.8%
- 41 GPT-4o Mini (snapshot unspecified) 27.5%
- 42 Llama-3.3-Nemotron 49B 25%
- 43 OLMo-3 7B Think 11.5%
- 44 Kimi-VL A3B 8.20%
- 45 Nemotron-3 30B 3.00%
Source: Official SLR-Bench leaderboard
- 81.8 GPT-5
- 80.8
Gemini 3 Flash
- 79.5 GPT-5 mini
- 67.5 gpt-oss-120b
- 66.0 GPT-5.4 Mini
- 65.2
Gemini 3.1 Flash-Lite
- 65.0
Gemini 2.5 Flash
- 62.2 o3-mini
- 61.5 o4-mini
- 60.0
Qwen3.5-27B
- 59.8 GPT-5 Chat
- 57.8
Qwen3.5-35B-A3B
- 55.0 GLM-4.5-Air
- 54.8
Gemma 4 31B IT
- 53.5 Kimi K2.5
- 52.8 Kimi K2
- 51.2 gpt-oss-20b
- 51.0
Llama-3.1-Nemotron-Ultra-253B-v1
- 51.0 GPT-5 nano
- 49.2
Nemotron-Cascade 2 30B-A3B
- 49.0
Qwen3.5-9B
- 47.8
Gemini 2.5 Flash-Lite
- 47.0
Ministral 3 14B
- 45.0
Olmo 3 32B Think
- 44.5
Qwen3.5-4B
- 43.8
DeepSeek-R1-Distill-Llama-70B
- 43.8
DeepSeek-R1-Distill-Qwen-32B
- 43.8
Gemma 4 26B A4B
- 41.2 GPT-4.5
- 40.8
Gemma 4 E4B
- 38.5
Olmo 3.1 32B Think
- 33.0
Gemma 4 E2B
- 1 GPT-5 medium 81.8%
- 2
Gemini 3 Flash high 80.8%
- 3 GPT-5 mini high 79.5%
- 4 gpt-oss-120b high 67.5%
- 5 GPT-5.4 Mini medium 66%
- 6
Gemini 3.1 Flash-Lite high 65.2%
- 7
Gemini 2.5 Flash Sep 2025 65%
- 8 o3-mini medium 62.2%
- 9 o4-mini medium 61.5%
- 10
Qwen3.5-27B 60%
- 11 GPT-5 Chat 59.8%
- 12
Qwen3.5-35B-A3B 57.8%
- 13 GLM-4.5-Air 55%
- 14
Gemma 4 31B IT 54.8%
- 15 Kimi K2.5 53.5%
- 16 Kimi K2 Sep 2025 52.8%
- 17 gpt-oss-20b high 51.2%
- 18
Llama-3.1-Nemotron-Ultra-253B-v1 51%
- 19 GPT-5 nano high 51%
- 20
Nemotron-Cascade 2 30B-A3B 49.2%
- 21
Qwen3.5-9B 49%
- 22
Gemini 2.5 Flash-Lite Sep 2025, high 47.8%
- 23
Ministral 3 14B 47%
- 24
Olmo 3 32B Think 45%
- 25
Qwen3.5-4B 44.5%
- 26
DeepSeek-R1-Distill-Llama-70B 43.8%
- 27
DeepSeek-R1-Distill-Qwen-32B 43.8%
- 28
Gemma 4 26B A4B 43.8%
- 29 GPT-4.5 41.2%
- 30
Gemma 4 E4B 40.8%
- 31
Olmo 3.1 32B Think 38.5%
- 32
Gemma 4 E2B 33%
Source: Official SLR-Bench leaderboard
| # | Entry | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-5 | 81.8% | medium | Aug 2025 | Apr 2026 | Official leaderboard |
| 2 | | 80.8% | high | Dec 2025 | Apr 2026 | Official leaderboard |
| 3 | GPT-5 mini | 79.5% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 4 | gpt-oss-120b | 67.5% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 5 | GPT-5.4 Mini | 66% | medium | Mar 2026 | Apr 2026 | Official leaderboard |
| 6 | GLM-4.7 FP8 | 65.5% | Apr 2026 | Official leaderboard | ||
| 7 | | 65.2% | high | Mar 2026 | Apr 2026 | Official leaderboard |
| 8 | | 65% | Sep 2025 | Apr 2026 | Official leaderboard | |
| 9 | o3-mini | 62.2% | medium | Jan 2025 | Apr 2026 | Official leaderboard |
| 10 | o4-mini | 61.5% | medium | Apr 2025 | Apr 2026 | Official leaderboard |
| 11 | | 60% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 12 | GPT-5 Chat | 59.8% | Aug 2025 | Apr 2026 | Official leaderboard | |
| 13 | Gemini-2.5-pro (snapshot unspecified) | 58% | Apr 2026 | Official leaderboard | ||
| 14 | | 57.8% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 15 | GLM-4.5-Air | 55% | Jul 2025 | Apr 2026 | Official leaderboard | |
| 16 | | 54.8% | Apr 2026 | Apr 2026 | Official leaderboard | |
| 17 | Kimi K2.5 | 53.5% | Jan 2026 | Apr 2026 | Official leaderboard | |
| 18 | Nemotron-3 120B | 53% | Apr 2026 | Official leaderboard | ||
| 19 | Kimi K2 Sep 2025 | 52.8% | Sep 2025 | Apr 2026 | Official leaderboard | |
| 20 | gpt-oss-20b | 51.2% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 21 | | 51% | Apr 2025 | Apr 2026 | Official leaderboard | |
| 22 | GPT-5 nano | 51% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 23 | | 49.2% | Mar 2026 | Apr 2026 | Official leaderboard | |
| 24 | | 49% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 25 | | 47.8% | high | Sep 2025 | Apr 2026 | Official leaderboard |
| 26 | | 47% | Dec 2025 | Apr 2026 | Official leaderboard | |
| 27 | | 45% | Nov 2025 | Apr 2026 | Official leaderboard | |
| 28 | | 44.5% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 29 | | 43.8% | Jan 2025 | Apr 2026 | Official leaderboard | |
| 30 | | 43.8% | Jan 2025 | Apr 2026 | Official leaderboard | |
| 31 | | 43.8% | Apr 2026 | Apr 2026 | Official leaderboard | |
| 32 | GPT-4.5 | 41.2% | Feb 2025 | Apr 2026 | Official leaderboard | |
| 33 | Ministral-3 8B | 41.2% | Apr 2026 | Official leaderboard | ||
| 34 | | 40.8% | Mar 2026 | Apr 2026 | Official leaderboard | |
| 35 | | 38.5% | Dec 2025 | Apr 2026 | Official leaderboard | |
| 36 | DeepSeek-R1-Distill Qwen 8B | 38.2% | Apr 2026 | Official leaderboard | ||
| 37 | Ministral-3 3B | 33.8% | Apr 2026 | Official leaderboard | ||
| 38 | | 33% | Mar 2026 | Apr 2026 | Official leaderboard | |
| 39 | GPT-4o (snapshot unspecified) | 32.2% | Apr 2026 | Official leaderboard | ||
| 40 | GPT-4 Turbo (snapshot unspecified) | 28.8% | Apr 2026 | Official leaderboard | ||
| 41 | GPT-4o Mini (snapshot unspecified) | 27.5% | Apr 2026 | Official leaderboard | ||
| 42 | Llama-3.3-Nemotron 49B | 25% | Apr 2026 | Official leaderboard | ||
| 43 | OLMo-3 7B Think | 11.5% | Apr 2026 | Official leaderboard | ||
| 44 | Kimi-VL A3B | 8.20% | Apr 2026 | Official leaderboard | ||
| 45 | Nemotron-3 30B | 3.00% | Apr 2026 | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-5 | 81.8% | medium | Aug 2025 | Apr 2026 | Official leaderboard |
| 2 | | 80.8% | high | Dec 2025 | Apr 2026 | Official leaderboard |
| 3 | GPT-5 mini | 79.5% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 4 | gpt-oss-120b | 67.5% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 5 | GPT-5.4 Mini | 66% | medium | Mar 2026 | Apr 2026 | Official leaderboard |
| 6 | | 65.2% | high | Mar 2026 | Apr 2026 | Official leaderboard |
| 7 | | 65% | Sep 2025 | Apr 2026 | Official leaderboard | |
| 8 | o3-mini | 62.2% | medium | Jan 2025 | Apr 2026 | Official leaderboard |
| 9 | o4-mini | 61.5% | medium | Apr 2025 | Apr 2026 | Official leaderboard |
| 10 | | 60% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 11 | GPT-5 Chat | 59.8% | Aug 2025 | Apr 2026 | Official leaderboard | |
| 12 | | 57.8% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 13 | GLM-4.5-Air | 55% | Jul 2025 | Apr 2026 | Official leaderboard | |
| 14 | | 54.8% | Apr 2026 | Apr 2026 | Official leaderboard | |
| 15 | Kimi K2.5 | 53.5% | Jan 2026 | Apr 2026 | Official leaderboard | |
| 16 | Kimi K2 Sep 2025 | 52.8% | Sep 2025 | Apr 2026 | Official leaderboard | |
| 17 | gpt-oss-20b | 51.2% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 18 | | 51% | Apr 2025 | Apr 2026 | Official leaderboard | |
| 19 | GPT-5 nano | 51% | high | Aug 2025 | Apr 2026 | Official leaderboard |
| 20 | | 49.2% | Mar 2026 | Apr 2026 | Official leaderboard | |
| 21 | | 49% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 22 | | 47.8% | high | Sep 2025 | Apr 2026 | Official leaderboard |
| 23 | | 47% | Dec 2025 | Apr 2026 | Official leaderboard | |
| 24 | | 45% | Nov 2025 | Apr 2026 | Official leaderboard | |
| 25 | | 44.5% | Feb 2026 | Apr 2026 | Official leaderboard | |
| 26 | | 43.8% | Jan 2025 | Apr 2026 | Official leaderboard | |
| 27 | | 43.8% | Jan 2025 | Apr 2026 | Official leaderboard | |
| 28 | | 43.8% | Apr 2026 | Apr 2026 | Official leaderboard | |
| 29 | GPT-4.5 | 41.2% | Feb 2025 | Apr 2026 | Official leaderboard | |
| 30 | | 40.8% | Mar 2026 | Apr 2026 | Official leaderboard | |
| 31 | | 38.5% | Dec 2025 | Apr 2026 | Official leaderboard | |
| 32 | | 33% | Mar 2026 | Apr 2026 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Basic, Easy and Medium. GPT-5 leads 2 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Basic
32 models we track have a score here.
- 1 o3-mini ↑7 100%
- 2 gpt-oss-120b ↑2 100%
- 3 GPT-5 ↓2 100%
- 4 GPT-5 Chat ↑7 100%
- 5 GPT-5 mini ↓2 100%
- 6 GPT-5 nano ↑13 100%
- 7
Gemini 2.5 Flash Sep 2025 100%
- 8
Gemini 3 Flash ↓6 100%
- 9
Qwen3.5-27B ↑1 100%
- 10
Qwen3.5-9B ↑11 100%
Easy
32 models we track have a score here.
- 1 GPT-5 100%
- 2 GPT-5 mini ↑1 100%
- 3
Gemini 3 Flash ↓1 98%
- 4 gpt-oss-120b 97%
- 5 o3-mini ↑3 96%
- 6
Gemini 2.5 Flash Sep 2025 ↑1 96%
- 7
Gemini 3.1 Flash-Lite ↓1 96%
- 8
Qwen3.5-27B ↑2 94%
- 9 o4-mini 93%
- 10
Qwen3.5-35B-A3B ↑2 92%
Medium
32 models we track have a score here.
- 1 GPT-5 77%
- 2 GPT-5 mini ↑1 74%
- 3
Gemini 3 Flash ↓1 71%
- 4 gpt-oss-120b 58%
- 5 GPT-5.4 Mini 52%
- 6
Gemini 3.1 Flash-Lite 48%
- 7
Gemini 2.5 Flash Sep 2025 46%
- 8
Gemma 4 31B IT ↑6 44%
- 9 o3-mini ↓1 43%
- 10 o4-mini ↓1 40%
Hard
32 models we track have a score here.
- 1
Gemini 3 Flash ↑1 54%
- 2 GPT-5 ↓1 50%
- 3 GPT-5 mini 44%
- 4 GPT-5.4 Mini ↑1 28%
- 5
Gemini 2.5 Flash Sep 2025 ↑2 18%
- 6
Gemini 3.1 Flash-Lite 18%
- 7 gpt-oss-120b ↓3 15%
- 8 o4-mini ↑1 14%
- 9 GPT-5 Chat ↑2 14%
- 10
Gemma 4 31B IT ↑4 12%
The best score, 81.8%, costs $103 for the full run
Runs on the cost frontier score higher than any run that's cheaper. GPT-5 mini at high comes within 2.3 points of the best for $28 for the full run, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official SLR-Bench leaderboard
| Model | Effort | Harness | Score | Cost of the full run | Output tokens |
|---|---|---|---|---|---|
| GPT-5 | medium | 81.8% | $103 | ||
| high | 80.8% | $81 | |||
| GPT-5 mini | high | 79.5% | $28 | ||
| GPT-5 mini | medium | 67.8% | $12 | ||
| gpt-oss-120b | high | 67.5% | $3.83 | ||
| GPT-5.4 Mini | medium | 66% | $81 | ||
| high | 65.2% | $24 | |||
| 65% | $58 | ||||
| o3-mini | medium | 62.2% | $75 | ||
| o4-mini | medium | 61.5% | $42 | ||
| 60% | $17 | ||||
| GPT-5 Chat | 59.8% | $36 | |||
| 57.8% | $15 | ||||
| GLM-4.5-Air | 55% | $8.01 | |||
| gpt-oss-120b | medium | 55% | $1.18 | ||
| GPT-5 mini | low | 54.8% | $4.07 | ||
| 54.8% | $3.87 | ||||
| Kimi K2.5 | 53.5% | $16 | |||
| Kimi K2 Sep 2025 | 52.8% | $20 | |||
| gpt-oss-20b | high | 51.2% | $4.17 | ||
| 51% | $4.98 | ||||
| GPT-5 nano | high | 51% | $6.62 | ||
| gpt-oss-20b | medium | 50% | $3.90 | ||
| 49.2% | |||||
| 49% | $2.07 | ||||
| high | 47.8% | $8.26 | |||
| low | 47.5% | $3.98 | |||
| medium | 47.2% | $4.57 | |||
| GPT-5 nano | medium | 47% | $2.81 | ||
| 47% | $0.82 | ||||
| 45% | $9.04 | ||||
| 44.5% | |||||
| gpt-oss-120b | low | 44.2% | $0.44 | ||
| 43.8% | $9.55 | ||||
| 43.8% | $4.32 | ||||
| 43.8% | $5.29 | ||||
| GPT-4.5 | 41.2% | $576 | |||
| 40.8% | |||||
| 38.5% | |||||
| GPT-5 nano | low | 37.8% | $0.80 | ||
| gpt-oss-20b | low | 37% | $0.37 | ||
| 33% | |||||
| 17.5% | $1.16 |
The best score, 81.8%, costs $103 for the full run
Runs on the cost frontier score higher than any run that's cheaper. GPT-5 mini at high comes within 2.3 points of the best for $28 for the full run, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official SLR-Bench leaderboard
| Model | Effort | Harness | Score | Cost of the full run | Output tokens |
|---|---|---|---|---|---|
| GPT-5 | medium | 81.8% | $103 | ||
| high | 80.8% | $81 | |||
| GPT-5 mini | high | 79.5% | $28 | ||
| GPT-5 mini | medium | 67.8% | $12 | ||
| gpt-oss-120b | high | 67.5% | $3.83 | ||
| GPT-5.4 Mini | medium | 66% | $81 | ||
| high | 65.2% | $24 | |||
| 65% | $58 | ||||
| o3-mini | medium | 62.2% | $75 | ||
| o4-mini | medium | 61.5% | $42 | ||
| 60% | $17 | ||||
| GPT-5 Chat | 59.8% | $36 | |||
| 57.8% | $15 | ||||
| GLM-4.5-Air | 55% | $8.01 | |||
| gpt-oss-120b | medium | 55% | $1.18 | ||
| GPT-5 mini | low | 54.8% | $4.07 | ||
| 54.8% | $3.87 | ||||
| Kimi K2.5 | 53.5% | $16 | |||
| Kimi K2 Sep 2025 | 52.8% | $20 | |||
| gpt-oss-20b | high | 51.2% | $4.17 | ||
| 51% | $4.98 | ||||
| GPT-5 nano | high | 51% | $6.62 | ||
| gpt-oss-20b | medium | 50% | $3.90 | ||
| 49.2% | |||||
| 49% | $2.07 | ||||
| high | 47.8% | $8.26 | |||
| low | 47.5% | $3.98 | |||
| medium | 47.2% | $4.57 | |||
| GPT-5 nano | medium | 47% | $2.81 | ||
| 47% | $0.82 | ||||
| 45% | $9.04 | ||||
| 44.5% | |||||
| gpt-oss-120b | low | 44.2% | $0.44 | ||
| 43.8% | $9.55 | ||||
| 43.8% | $4.32 | ||||
| 43.8% | $5.29 | ||||
| GPT-4.5 | 41.2% | $576 | |||
| 40.8% | |||||
| 38.5% | |||||
| GPT-5 nano | low | 37.8% | $0.80 | ||
| gpt-oss-20b | low | 37% | $0.37 | ||
| 33% | |||||
| 17.5% | $1.16 |
3 models have held number one on the official leaderboard since its release in June 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official SLR-Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official SLR-Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Carbon 97%
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels.
Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (45 entries).
- Pass Models scored 32 models we track, plus 13 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Flash (#60).
- Pass Headroom The best is 81.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 43 rows for models we track and 13 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.