whatmodel.sh / GPT-5.6 Terra

GPT-5.6 Terra

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

GPT-5.6 Terra is a closed-weights language model from OpenAI, released July 2026.

78.5±0.5#17 of 719 LLMsfrom 38 benchmark results across 11 sources

$ how it compares

The LLMs ranked either side of GPT-5.6 Terra, out of 828 with a score, then more from OpenAI.

#modelscorewhereRAMreleased
15MiMo-V2.6-Pro Xiaomi79±1api $0.54/M +local~443GB2026-09
16Hy4 preview (770B-A49B) Tencent79±1api $1.25/M +local~248GB2026-08
17GPT-5.6 Terra OpenAI79±1api $4.50/M2026-07
18Qwen3.8 Max (2.4T-A95B) Alibaba78±1api $3.00/M +local~380GB2026-08
19Claude Opus 4.8 Anthropic78±0api $10.00/M2026-05
more from OpenAI
1GPT-6 Astra OpenAI87±1api $20.00/M2026-09
3GPT-6.1 Sol OpenAI85±1api $4.00/M2026-09
7GPT-6 Sol OpenAI83±1api $4.00/M2026-09

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #26 of 300 · 23% of its score · 10 tests

Trails GLM-5.3-Flash, Grok 4.7, Qwen3.8 Max (0902) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena Agent (October 2026 aggregate IPS)0.0040 IPS ✎▁▁▁▄▂▃▇█▂▂▁▃ beats 20/49avg -0.312 IPS▇▇ 5.1%sharp
Terminal-Bench 4.021.5% ✎█▅▅▅▅·▅▃▃▅▅▅ beats 8/20very hard · avg 0.4%▇ 3.3%sharp · at its level
BrowseComp lab87.5% ✎▄··█▄··▄·▄█▄ beats 4/8hard · avg 32%▇ 2.5%typical
SWE-bench Pro (split unspecified) lab63.4% ✎▃·▁▁▁▂▂▃█▇▃▁ beats 45/55very hard · avg 5.1%▇ 2.4%sharp · at its level
Terminal-Bench 2.1 lab87.4% ✎▁·▁·▁▁▄▃▄█▃▇ beats 50/56hard · avg 21%▇ 2.4%typical
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@151.8% ✎▃··▅··▃█▃▅▃▅ beats 8/12very hard · avg 3.1%▇ 2.0%sharp · at its level
Toolathlon (Tool Decathlon) lab53.1% ✎▃··▃··▃··▃█▅ beats 4/8very hard · avg 13%▇ 1.8%at its level
DeepSWE v1.1 (resolved) lab69.6% ✎█··▃▃····▃·▃ beats 4/6hard · avg 21%▇ 1.4%at its level
2 more tests
testresultvs the others in its rundifficultycountswhy
Agents' Last Exam (metric/harness unspecified) lab50.4% ✎▂·▂·█▂···▂▄▄ beats 9/11very hard · avg 11%▇ 1.0%at its level
KlinikeBench strict pass@1 (333 interactive clinical encounters, Terminus-2)8.4% ✎▁▁·▁▁▁▁▅▃█▅█ beats 8/27very hard · avg 2.4%▇ 0.8%typical
coding▁▁▁▃▆██▇▆▄▃▁ #43 of 458 · 12% of its score · 5 tests

Beats GPT-5.5 here, though they rank higher overall. Trails Mistral Large 4 (public preview), Qwen3.8 27B, GLM-5.2 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena WebDev (Code Arena)1519.1 ✎▁▂▄▃▅█▄▇▆▄▁▁ beats 77/119avg 981▇▇ 4.6%sharp
SWE-bench Pro (split unspecified) lab63.4% ✎▃·▁▁▁▂▂▃█▇▃▁ beats 45/55very hard · avg 5.1%▇ 2.4%sharp · at its level
E2E-SWE (build a working codebase from scratch, 186 tasks) pass@151.8% ✎▃··▅··▃█▃▅▃▅ beats 8/12very hard · avg 3.1%▇ 2.0%sharp · at its level
Luau Leetcode fully solved (Torpedo Software)81.3% ✎▁▁▁▁·▁▂▃▅▃█▇ beats 36/49hard · avg 18%▇ 1.5%typical
DeepSWE v1.1 (resolved) lab69.6% ✎█··▃▃····▃·▃ beats 4/6hard · avg 21%▇ 1.4%at its level
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #36 of 393 * · 2.6% of its score · 1 test

Beats GPT-6 Sol here, though they rank higher overall. Trails MiMo-V2.5-Pro (1T-A42B), Grok 4.20 (0309 reasoning), Claude Opus 4.5 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1465.5 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 337/379avg 1268▇ 2.6%blunt
math▁▂▃▅▆▅▄█▄▁▃▁ #13 of 398 · 38% of its score · 3 tests

Beats GPT-5.4 Pro here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
FrontierMath tier 4 v2 +170.7% ✎▁▂▁▂█▅▂▃▁▁▁▁ beats 52/63very hard · avg 0.1%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level · several runs
FrontierMath tiers 1–3 v2 +186.0% ✎▂··▁▁▃▄▄█▅▃▃ beats 73/82very hard · avg 1.2%▇▇▇▇▇▇ 15%sharp · several runs
OTIS mock AIME 2024–202599.7% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 190/193very hard · avg 8.5%▇ 0.2%sharp · maxed out · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #30 of 147 · 5.1% of its score · 4 tests

Beats Claude Sonnet 5.5, MiMo-V2.6-Pro here, though they rank higher overall. Trails Qwen3.6 Plus, Qwen3.7 Plus, Gemini 3 Flash (Preview) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena vision (style controlled)1268.2 ✎▁▁▂▂▂▂▄▅▅▆██ beats 116/145avg 1068▇ 1.8%blunt
MMMU Pro lab80.7% ✎▅·····▃·▃·█▃ beats 4/7hard · avg 40%▇ 1.6%typical
Arena document (style controlled)1476.7 ✎▁▁▃▃▄▃▇█▅▇▄▅ beats 27/37avg 1335▇ 0.9%blunt
SciDocBench scientific-document workflows (overall, 496 instances)54.2% ✎▃▃···█▃▅▃·▃▅ beats 7/11hard · avg 26%▇ 0.8%blunt · at its level
web search▂▁▂▆▄█▃▃▁▁·▁ #13 of 80 * · 2.5% of its score · 1 test

Beats Claude Fable 5.1, Claude Sonnet 5.5, Claude Opus 5 here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), GPT-4o (2024-08-06) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BrowseComp lab87.5% ✎▄··█▄··▄·▄█▄ beats 4/8hard · avg 32%▇ 2.5%typical
other tests16 not in a task · 25% of its score
testresultvs the others in its rundifficultycountswhy
Furniture Assembly (Epoch)54.2% ✎█▆▆▆▆▂▅▂▂▂·▅ beats 22/30very hard · avg 1.2%▇▇ 4.7%sharp · at its level
Mystery Game Puzzles (Epoch)35.0% ✎█··▂▃▇▄█▂▃▁▁ beats 57/74very hard · avg 9.4%▇▇ 4.5%sharp · at its level · noisy
KMMLU-HARD (Daekeun Kim, zero-shot)77.9% ✎█···▁▁▁▂▃·▆▂ beats 21/23very hard · avg 29%▇ 2.6%sharp
Chess Puzzles (Epoch)54.0% ✎█▁▁▁▁▂▁▄▂▂▃▁ beats 130/135very hard · avg 5.9%▇ 2.5%sharp · at its level · noisy
GPQA Diamond93.3% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 196/208hard · avg 40%▇ 1.3%sharp · maxed out
KoBALT-700 (Daekeun Kim)76.6% ✎▄▄█·····▄··▄ beats 2/5hard · avg 37%▇ 1.3%typical
KMMLU-Pro (Daekeun Kim)84.4% ✎▄·▄█·····▄·▄ beats 1/5medium · avg 40%▇ 1.1%typical
CLIcK Korean (Daekeun Kim)92.2% ✎▆▄▄▄█▄▂·▄█·▄ beats 21/23medium · avg 58%▇ 1.1%maxed out
8 more tests
testresultvs the others in its rundifficultycountswhy
SimpleQA Verified43.2% ✎▁▁▃▂▃▃▆▆█▃▂▅ beats 39/78very hard · avg 8.0%▇ 1.0%at its level
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)53.2% ✎▁▁▃▂▂▄▁▃▄▄▄█ beats 22/38very hard · avg 14%▇ 0.9%at its level
AutomationBench public set (Zapier README)37.2% ✎▄······▄█··▄ beats 0/4very hard · avg 10%▇ 0.9%at its level
HAE-RAE 1.0 (Daekeun Kim)92.8% ✎▄▄▄▂▂█▂▄▂█▂▄ beats 19/22medium · avg 63%▇ 0.8%maxed out
MuSR Korean (Daekeun Kim)83.2% ✎█·····███·██ beats 3/5medium · avg 42%▇ 0.7%typical
HealthBench Hard lab34.3% ✎▂··▂····▂▄·█ beats 8/8very hard · avg 9.7%▇ 0.6%blunt · at its level
HealthBench lab58.7% ✎▃·▃·····█▃▃▅ beats 8/8hard · avg 24%▇ 0.6%blunt · at its level
RobloxQA v2.0 multiple choice (Torpedo Software)95.8% ✎▁▁▃▁█▄▃▃▆▂▂▂ beats 31/49easy · avg 78%▇ 0.5%maxed out

$ where to run it

Through an API

api · $4.50/M · $2 in / $12 out per 1M tokens · 1.1M context · platform.openai.com ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

$ how the score was worked out · Epoch AI — internal evaluatio… 35%, OpenAI vendor-reported benchm… 31%, LMArena leaderboard dataset 15% and 8 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Epoch AI — internal evaluations8▇▇▇▇▇▇▇▇▇▇▇ 35%CC BY 4.0
OpenAI vendor-reported benchmark scores for its own models (openai.com announcement posts and Deployment Safety Hub system cards) · lab-reported11▇▇▇▇▇▇▇▇▇ 31%Factual vendor-reported score (not licensed content)
LMArena leaderboard dataset5▇▇▇▇ 15%CC BY 4.0
Daekeun Kim Korean LLM benchmark runs6▇▇ 8%MIT
Terminal-Bench 4.0 official leaderboard submissions1▇ 3%Apache-2.0
E2E-SWE: Benchmarking LLMs on Building Working Codebases from Scratch (arXiv:2609.38335v1)1▇ 2%CC BY 4.0
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode2▇ 2%MIT
Miðeind Icelandic LLM Leaderboard1▇ 1%CC BY 4.0
Zapier AutomationBench README: public-set pass rates1▇ 1%MIT
SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding (arXiv:2609.05141v2)1▇ 1%CC BY 4.0
KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy (arXiv:2609.38480v2)1▇ 1%CC BY 4.0
prior · pull toward the middle▇ 0%
weighted average= 78.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json