GPT-6 Sol benchmarks, price, and specs | Tokens or Towers

AA Intelligence: 47.5#7 of 142
Input / output per 1M: $2.00/ $10.00
Output speed: 87tok/s
Context window: 1.05M tokens

Composite indexes

Artificial Analysis Intelligence Index#7 of 142
48independentAA Intelligence Index v4.3.2, max effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Opus 5.5 at 58

Coding

Terminal-Bench 4.0#7 of 88
43.9%independentTerminal-Bench 4.0 (AA harness), max effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Mythos 5.1 at 60.9%
SciCode#12 of 94
57.6%independentmax effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Opus 5.5 at 66.9%

Agents and tool use

GDPval-AA#20 of 103
1487independentGDPval-AA, max effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Opus 5.5 at 1846
OSWorld-Verified#14 of 16
60.5%vendorOSWorld 2.0 offline, xhighOpenAI: Introducing GPT-6 Sol and Luna ↗
Leader: Kimi K3 at 84.8%

Reasoning and knowledge

Humanity’s Last Exam#11 of 147
47.9%independentno tools (AA), max effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Opus 5.5 at 61.4%
CritPt#4 of 139
30.9%independentmax effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: GPT-5.6 Sol at 32.3%

Vision and long context

AA Long Context Reasoning#8 of 139
83.7%independentmax effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Kimi K3 at 88.7%
MMMU-Pro#8 of 72
83.3%independentAA, max effortArtificial Analysis: GPT-6 Sol (max) ↗
Leader: Claude Opus 5.5 at 87.7%

Human preference

LMArena WebDev#5 of 54
1686independentgpt-6-sol-max, LMArena Code arena (formerly WebDev), vote cutoff 2026-09-23LMArena Code (WebDev) leaderboard ↗
Leader: Claude Opus 5.5 at 1818

Other published results

Result	Score	Reported by	Source
AA-Omniscience Index · max effort	27.1 index (-100 to 100)	independent	Artificial Analysis: GPT-6 Sol (max) ↗
AutomationBench-AA (partial score) · max effort	61.6%	independent	Artificial Analysis: GPT-6 Sol (max) ↗
Terminal-Bench Science (AA) · max effort	30%	independent	Artificial Analysis: GPT-6 Sol (max) ↗
AA Coding Agent Index v1.5 · Codex + GPT-6 Sol (max); DeepSWE 69.0 / SWE-Atlas-QnA 57.5 / TB4 43.4	56.7 index	independent	Artificial Analysis Coding Agents ↗
Agents' Last Exam · max	56.4%	vendor	OpenAI: Introducing GPT-6 Sol and Luna ↗
DeepSWE v1.1 · max	68.8%	vendor	OpenAI: Introducing GPT-6 Sol and Luna ↗
AutomationBench · xhigh ($0.27/task)	33.2%	vendor	OpenAI: Introducing GPT-6 Sol and Luna ↗