OpenAImodel systemcard
GPT-5.6 Sol
#3of 27 · general
xhigh/max effort collapsed
1.1Mcontext
Proprietarynon-commercial
2026-07-09released
2026-02-16knowledge cutoff
Standing
dot = this model · shaded = the field's middle 60% · track = full fieldGeneral
85.2#3/27
Math
81.1#3/27
Coding
90.0#1/27
Health
54.9#11/26
Engineering
87.5#3/17
Cyber
100.0#1/14
Analyst
70.0#4/11
Legal
55.4#8/26
Instruct
70.4#4/27
Agents
64.9#7/27
Professional
79.3#5/27
Multimodal
68.0#3/19
Hallucination
49.4#7/27
Speed
63.8#6/22
Value
66.4#4/27
Measured
≈ imputed · amber dot = lab-self-reported among the sourcesGPQA Diamond94.4%#1
HLE49.5%#5
MMLU-Pro89.1%#8
LiveBench81.0%#1
SWE-bench V96.2%#2
LiveCodeBench82.6%#16
SciCode56.9%#5
CritPt32.3%#1
Terminal-Bench87.6%#1
APEX Agents39.9%#4
GDPval61.4%#4
Arena1485#6
AA Index60.9#3
Vals Index73.1#4
Epoch Capabilities Index161.7#1
Operations
$5 / $30$ per 1M tokens, in / out
$11blended
64tokens / sec
$0.44$ per solved task
1485arena elo · 8,359 votes
Series
OpenAI's models on the board, by releaseGPT-5.52026-04-23 · 67GPT-5.6 Sol2026-07-09 · 85GPT-5.6 Terra2026-07-09 · 59GPT-5.6 Luna2026-07-09 · 50