benchgap
Calibration

Terminal-Bench 2.1 → AA Terminal-Bench 4.0

AA Terminal-Bench 4.0 is estimated from Terminal-Bench 2.1 with a Hill curve fitted on 7 models measured on both: y = 0.0000 + (1.2000 − 0.0000)·x^6.00 / (1.06990^6.00 + x^6.00), R² = 0.62, cross-validated error 11.6 pp. It is used for 64 estimates.

Estimated modelTerminal-Bench 2.1AA Terminal-Bench 4.0Source
GPT-5.6 Sol91.9%34.4%estimated ± 11.6 pp, low confidence
Muse Spark 1.388.8%29.6%estimated ± 11.6 pp, low confidence
Kimi K388.3%28.8%estimated ± 11.6 pp, low confidence
Claude Mythos 588.0%28.4%estimated ± 11.6 pp, low confidence
DeepSeek V4 Pro 081387.9%28.2%estimated ± 11.6 pp, low confidence
GPT-5.6 Terra87.4%27.5%estimated ± 11.6 pp, low confidence
Qwen3.8 Max86.6%26.3%estimated ± 11.6 pp, low confidence
Ornith-1.5-397B86.1%25.6%estimated ± 11.6 pp, low confidence
Gemini 3.7 Flash85.8%25.2%estimated ± 11.6 pp, low confidence
Hy4 preview85.4%24.7%estimated ± 11.6 pp, low confidence
GPT-5.6 Luna84.7%23.7%estimated ± 11.6 pp, low confidence
Claude Fable 584.3%23.2%estimated ± 11.6 pp, low confidence
GLM-5.3-Flash84.3%23.2%estimated ± 11.6 pp, low confidence
Grok 4.583.3%21.9%estimated ± 11.6 pp, low confidence
Ultramarine 1 Flash83.2%21.7%estimated ± 11.6 pp, low confidence
Muse Spark 1.282.9%21.3%estimated ± 11.6 pp, low confidence
DeepSeek V4 Flash 073182.7%21.1%estimated ± 11.6 pp, low confidence
Sakana Fugu-Ultra82.1%20.3%estimated ± 11.6 pp, low confidence
Ember-182.0%20.2%estimated ± 11.6 pp, low confidence
SWE-1.781.5%19.6%estimated ± 11.6 pp, low confidence
GLM-5.281.0%19.0%estimated ± 11.6 pp, low confidence
Claude Sonnet 580.4%18.3%estimated ± 11.6 pp, low confidence
Sakana Fugu80.2%18.1%estimated ± 11.6 pp, low confidence
Muse Spark 1.180.0%17.9%estimated ± 11.6 pp, low confidence
Particle-179.9%17.8%estimated ± 11.6 pp, low confidence
Beam 1.0 501B-A23B FP878.9%16.6%estimated ± 11.6 pp, low confidence
Beam 1.0 501B-A23B NVFP478.4%16.1%estimated ± 11.6 pp, low confidence
Atria Dawn Preview78.3%16.0%estimated ± 11.6 pp, low confidence
Ornith-1.0-397B77.5%15.1%estimated ± 11.6 pp, low confidence
Gemini 3.5 Flash76.2%13.9%estimated ± 11.6 pp, low confidence
dots3-note Preview75.1%12.8%estimated ± 11.6 pp, low confidence
Claude Opus 4.874.6%12.4%estimated ± 11.6 pp, low confidence
Jais-3-375B-A23B73.8%11.7%estimated ± 11.6 pp, low confidence
OrcaCyber-Nano-1.072.5%10.6%estimated ± 11.6 pp, low confidence
Seed 2.1 Pro71.0%9.4%estimated ± 11.6 pp, low confidence
Apodex 1.170.8%9.3%estimated ± 11.6 pp, low confidence
Laguna S 2.170.2%8.9%estimated ± 11.6 pp, low confidence
Quasar 438B69.3%8.3%estimated ± 11.6 pp, low confidence
Ornith-1.5-35B-A3B67.8%7.3%estimated ± 11.6 pp, low confidence
Seed 2.1 Turbo67.6%7.2%estimated ± 11.6 pp, low confidence
Jais-3-MoVA-36B-A4B67.3%7.0%estimated ± 11.6 pp, low confidence
Nemotron 3.5 Super VL 120B A12B66.7%6.7%estimated ± 11.6 pp, low confidence
MiniMax M366.0%6.3%estimated ± 11.6 pp, low confidence
Pokee-Isaac 28B65.1%5.8%estimated ± 11.6 pp, low confidence
Inkling-Small64.7%5.6%estimated ± 11.6 pp, low confidence
Ornith-1.0-35B64.2%5.4%estimated ± 11.6 pp, low confidence
Jais-3-32B63.7%5.1%estimated ± 11.6 pp, low confidence
MAI-Code-1.1-Flash62.9%4.8%estimated ± 11.6 pp, low confidence
Nemotron 3.5 Super VL 120B A12B NVFP460.8%3.9%estimated ± 11.6 pp, low confidence
Step 3.7 Flash59.5%3.4%estimated ± 11.6 pp, low confidence
Ling 3.0 Flash57.0%2.7%estimated ± 11.6 pp, low confidence
Solar Pro 457.0%2.7%estimated ± 11.6 pp, low confidence
Nemotron 3 Ultra56.4%2.5%estimated ± 11.6 pp, low confidence
Flint54.9%2.2%estimated ± 11.6 pp, low confidence
Gemini 3.5 Flash-Lite54.0%2.0%estimated ± 11.6 pp, low confidence
Nemotron 3.5 Bolt 74B-A9B BF1653.9%1.9%estimated ± 11.6 pp, low confidence
Ternary Bonsai 2 27B52.8%1.7%estimated ± 11.6 pp, low confidence
Muse Glimmer 30B51.7%1.5%estimated ± 11.6 pp, low confidence
Nemotron 3.5 Bolt 74B-A9B51.0%1.4%estimated ± 11.6 pp, low confidence
Jais-3-7B46.4%0.8%estimated ± 11.6 pp, low confidence
Ornith-1.5-9B46.2%0.8%estimated ± 11.6 pp, low confidence
K-EXAONE 2.043.8%0.6%estimated ± 11.6 pp, low confidence
Ornith-1.0-9B43.1%0.5%estimated ± 11.6 pp, low confidence
Jais-3-3.7B41.2%0.4%estimated ± 11.6 pp, low confidence

Fitted on

ModelTerminal-Bench 2.1AA Terminal-Bench 4.0
Gemini 3.8 Flash89.4%19.7%
GLM-5.388.2%41.9%
Inkling63.8%1.0%
MiMo-V2.6-Pro89.9%34.8%
Qwen3.8-27B73.0%5.6%
Beam 1.0 501B-A23B80.1%3.0%
Step 5 Preview85.0%33.3%