Claude Sonnet 5.5 Scores 70.6% on Terminal-Bench 4.0, Beating Opus 5.5 at Half the Price
Anthropic has released Claude Sonnet 5.5, the second model in the Claude 5.5 family. The headline result: it scores 70.6% on Terminal-Bench 4.0, 4.2 percentage points above Opus 5.5’s 66.4% on the same agentic coding evaluation, while priced at $2 per million input tokens and $10 per million output tokens — half of Opus 5.5’s $4/$20.
Pricing is identical to Sonnet 5, but the model needs significantly fewer tokens to complete comparable work. Anthropic puts the cost-per-task reduction at up to 30% versus Sonnet 5.
Benchmark Table
| Evaluation | Sonnet 5.5 | Sonnet 5 | Opus 5.5 | GPT-6 Sol |
|---|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% | 66.4% | — |
| FrontierCode 1.1 Main | 46.2% Max / 52.1% Xhigh | 42.4% | 54.4% | 49.3% |
| CursorBench 4.0 | 55.5% | 34.1% | 57.8% | — |
| GDPval-AA v2.1 | 1844 | 1449 | 1846 | 1487 |
| AA-Briefcase v1.1 | 1811 | 1359 | 1822 | 1483 |
| HLE with tools | 64.5% | 54.9% | 67.7% | — |
| OSWorld 2.1 (partial) | 80.1% | 57.0% | 81.8% | — |
| Chartography | 61.6% | 15.6% | 64.4% | 53.6% |
The gap between Sonnet 5.5 and Sonnet 5 on Terminal-Bench 4.0 — 60 percentage points — is the largest of any benchmark in the comparison. The model closes nearly all the distance to Opus 5.5 on knowledge-work benchmarks: GDPval-AA v2.1 (1844 vs 1846) and AA-Briefcase v1.1 (1811 vs 1822).
Speed and Design
Sonnet 5.5 generates outputs 30%+ faster than Sonnet 5. Anthropic positions it for well-scoped everyday tasks: fixing bugs, creating polished documents, slides, and spreadsheets. Early testers described it as a better collaboration partner than Sonnet 5, with clearer natural-language communication during long-horizon work.
It is the first Sonnet model to beat Pokémon Red working only from screenshots — a proxy Anthropic uses to measure sustained visual reasoning and task persistence.
Safety and Availability
Because Sonnet 5.5’s cybersecurity capabilities are comparable to Opus 5’s, it launches with the same cyber safeguards and fallbacks applied to Anthropic’s most capable models. Biology safeguards match Sonnet 5’s. Both target a narrow set of high-risk requests and leave routine software development and most life sciences work unaffected.
Cache reads are priced at $0.20 per million tokens, matching Sonnet 5. Zero data retention is available. Haiku 5.5, designed for high-volume and cost-sensitive workloads, is coming in the weeks ahead.
Sonnet 5.5 is available now on the Claude API and across Amazon Web Services, Google Cloud, and Microsoft Azure. It is also generally available in GitHub Copilot.