GPT-5.6 Sol leads on the shared-benchmark average
Ahead on 8 of 11 shared benchmarks, averaging 7.8 points higher
Summarised only from the 11 percentage-scale benchmarks scored by every selected model; details are below. A further 2 Elo/rating-scale benchmarks are left out of the average — their scale cannot be added to percentages.
“Best available” takes each model’s highest recorded non-parallel mode per benchmark, so it may combine modes into a virtual configuration that does not exist. Read it with the mode breakdown.

GPT-5.6 Sol
OpenAI
Benchmark-by-benchmark comparison. Changing the thinking mode or tool filters updates the chart and table below.
“Best available” picks the highest non-parallel mode separately for each benchmark. The resulting series can combine several reasoning levels and is not one reproducible runtime configuration. Choose a mode filter to compare like-for-like runs.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Each axis is the mean percentage score of one benchmark domain. It is an average, not a capability rating.
Relative edge: 数学推理 +18.0 / Relative gap: 综合评估 -4.3
Relative edge: 综合评估 +4.3 / Relative gap: 数学推理 -18.0
Method: for each model and benchmark, all scores in the current mode scope are averaged (not the best score), then those benchmark scores are averaged within each domain. Only benchmarks scored on a 0-100 scale by at least two of the selected models count — Elo and rating-scale benchmarks such as Codeforces or Arena are excluded, because averaging a 1500 rating with an 85% accuracy produces a meaningless number. Missing values are not counted as zero, and the averages are unweighted, so domains with harder benchmarks read lower.
Every model and runtime mode, benchmark by benchmark. Values are comparable along a row, not between different benchmarks.
13 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.
| Benchmark | GPT-5.6 Sol | Claude Opus 4.8 |
|---|---|---|
HLE 综合评估 | 49.50Thinking Level · High | 57.90Extended Thinking | Tools |
GPQA Diamond 科学与综合推理 | 93.50Thinking Level · High | 93.60Thinking Level · High |
Creative Writing 写作和创作 | 1964.10Standard Mode | 1835.30Standard Mode |
DeepSWE 编程与软件工程 | 72.70Thinking Level · Extra High | Tools | 59.00Deep Thinking Mode | Tools |
SWE-Bench Pro - Public 编程与软件工程 | 64.60Thinking Level · Extra High | Tools | 69.20Extended Thinking | Tools |
Text Arena (Coding) 编程与软件工程 | 1620.27Thinking Level · Extra High | 1545.05Standard Mode |
WeirdML v2 编程与软件工程 | 88.76Thinking Level · High | Tools | 82.89Thinking Level · Extra High | Tools |
Context Arena 文本向量检索 | 97.63Thinking Level · High | 90.04Thinking Level · High |
Terminal-Bench 2.1 AI Agent - 工具使用 | 88.80Thinking Level · High | 78.90Thinking Level · High | Tools |
Terminal-Bench 4.0 AI Agent - 工具使用 | 37.27Thinking Level · High | Tools | 23.64Thinking Level · High | Tools |
Terminal-Bench-Science 0.1 AI Agent - 工具使用 | 22.40Thinking Level · High | Tools | 10.50Thinking Level · High | Tools |
82.93Thinking Level · High | 56.10Thinking Level · High | |
FrontierMath v2 数学推理 | 89.12Thinking Level · High | 80.00Thinking Level · High |
Official list prices per model API, split by input and output. Unit: USD per 1M tokens.
Architecture, licensing and API modalities. "Not provided" means the field is missing from our database.
| Features & specs | GPT-5.6 SolOpenAI | Claude Opus 4.8Anthropic |
|---|---|---|
Core specsRelease | 2026-06-26 | 2026-05-28 |
Context length | 1.05M | 1M |
Max output length | 128,000 tokens | 128,000 tokens |
Architecture | Undisclosed | Undisclosed |
Runtime modes | 关闭低中高极高最高 | 关闭低中高极高最高 |
Availability & licensingCode availability | Not public | Not public |
Weight availability | Not public | Not public |
Use & commercial terms | Official service only; subject to provider terms | Official service only; subject to provider terms |
API modality supportText Input/Output | Input:YesOutput:Yes | Input:YesOutput:Yes |
Image Input/Output | Input:YesOutput:No | Input:YesOutput:No |
ResourcesPaper / report | Previewing GPT-5.6 Sol: a next-generation model | Introducing Claude Opus 4.8 |
DataLearner blog | Not provided | Anthropic发布Claude Opus 4.8:定价不变,编程与智能体能力小幅提升, |

Claude Opus 4.8
Anthropic