Grouped by capability, sorted by largest gap within each. 12 shared benchmarks.
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| ARC-AGI-2 | 851 / 58极高强度思考（无工具） | 77.107 / 58Normal (No Tools) | +7.90 |
| ARC-AGI | 953 / 65极高强度思考（无工具） | 93.707 / 65Normal (No Tools) | +1.30 |
| GPQA Diamond | 93.606 / 175Thinking High (No Tools) | 92.809 / 175极高强度思考（无工具） | +0.80 |
| HLE | 52.2010 / 149Thinking High (With Tools) | 52.1011 / 149极高强度思考（工具） | +0.10 |
| ARC-AGI-3 | 02 / 6Thinking High (No Tools) | 04 / 6Thinking High (No Tools) | — |
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| Terminal Bench 2.0 | 82.701 / 43Thinking High (With Tools) | 75.104 / 43极高强度思考（工具） | +7.60 |
| OSWorld-Verified | 78.702 / 14Thinking High (With Tools) | 754 / 14极高强度思考（工具） | +3.70 |
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| FrontierMath - Tier 4 | 35.407 / 80Thinking High (With Tools) | 27.1011 / 80极高强度思考（无工具） | +8.30 |
| FrontierMath | 51.702 / 60Thinking High (With Tools) | 47.605 / 60极高强度思考（无工具） | +4.10 |
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 985 / 35Thinking High (With Tools) | 64.3030 / 35Normal (With Tools) | +33.70 |
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| BrowseComp | 84.405 / 43Thinking High (With Tools + Internet) | 82.709 / 43极高强度思考（工具） | +1.70 |
| Benchmark | GPT-5.5 | GPT-5.4 | Diff |
|---|---|---|---|
| SWE-Bench Pro - Public | 58.603 / 36Thinking High (With Tools) | 57.706 / 36极高强度思考（无工具） | +0.90 |
| Field | GPT-5.5 | GPT-5.4 |
|---|---|---|
| Publisher | OpenAI | OpenAI |
| Release date | 2026-04-23 | 2026-03-05 |
| Model type | Reasoning model | Multimodal model |
| Architecture | Dense | Dense |
| Parameters | 0.0 | 0.0 |
| Context length | 1000K | 1M |
| Max output | 131072 | 128000 |
Prices use DataLearner records when available; missing fields are not inferred.
| Item | GPT-5.5 | GPT-5.4 |
|---|---|---|
| Text input | $0.5 / 1M tokens | $2.5 / 1M tokens |
| Text output | $30 / 1M tokens | $15 / 1M tokens |
| Cache read | $0.5 / 1M tokens | Not public |
| Cache write | $6.25 / 1M tokens | $0.25 / 1M tokens |
On average across the 12 shared benchmarks, GPT-5.5 scores 5.84 higher.
Largest single-benchmark gap: τ²-Bench - Telecom — GPT-5.5 98 vs GPT-5.4 64.30 (+33.70).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.
