Claude Sonnet 5 vs GPT-5.6
Pricing, benchmarks, and use case comparison
Newer versions are out: Claude Sonnet 5 has been replaced by Claude Sonnet 5.5; GPT-5.6 has been replaced by GPT-6.1 Sol.
Verdict
Both are previous-generation models, replaced since September 2026 by Claude Sonnet 5.5 and OpenAI's GPT-6 family. Between the two, pick Claude Sonnet 5 for a flat $2/$10 price across its full 1M context, or GPT-5.6 if you route work across its Sol, Terra and Luna tiers, which all score higher on Terminal-Bench 2.1.
Specs comparison
| Claude Sonnet 5 | GPT-5.6 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Type | Closed source | Closed source |
| Context window | 1M | ✓1.05M |
| Input / 1M tokens | ✓$2.00 | $4.00 |
| Output / 1M tokens | $10.00 | $20.00 |
| Release date | 2026-06 | 2026-07 |
Benchmarks
| Benchmark | Claude Sonnet 5 | GPT-5.6 |
|---|---|---|
| SWE-bench Verified | ~85.2% | - |
| Terminal-Bench 2.1 | 80.4% | ✓91.9% |
| SWE-bench Pro | 63.2% | - |
| OSWorld-Verified | 81.2% | - |
| Terminal-Bench 2.1 (Sol) | - | 88.8% |
| Terminal-Bench 2.1 (Terra) | - | 82.5% |
| Terminal-Bench 2.1 (Luna) | - | 84.3% |
| ExploitBench | - | Competitive with Anthropic Mythos Preview |
| ExploitGym | - | Significant improvements |
| GeneBench v1 | - | Not announced |
Scores sourced from official provider release posts and independent benchmark aggregators.
Both models have been superseded: Claude Sonnet 5's successor is Claude Sonnet 5.5, and GPT-5.6's is OpenAI's GPT-6 family, including GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna.
Capability and benchmarks
Claude Sonnet 5 scores 80.4% on Terminal-Bench 2.1, around 85.2% on SWE-bench Verified, and 63.2% on SWE-bench Pro. The GPT-5.6 family spans three tiers on Terminal-Bench 2.1: Sol at 88.8%, Luna at 84.3%, and Terra at 82.5%. All three tiers edge Sonnet 5 on that benchmark, while Sonnet 5 has the broader set of published coding results. On our capability index Sonnet 5 rates coding 92 and reasoning 90, against 94 and 94 for the GPT-5.6 family.
Availability, price, and context
The gated preview that limited GPT-5.6 at launch ended when OpenAI made the family generally available across ChatGPT, Codex and the API on July 9, 2026. Both are still callable: Anthropic lists Sonnet 5 as a legacy model with retirement not sooner than June 30, 2027, and OpenAI has not announced a deprecation for GPT-5.6. Sonnet 5 costs $2/$10 per 1M tokens; Anthropic made its introductory price permanent and cancelled the planned rise to $3/$15. Its 1M-token context is billed at the standard rate. OpenAI's pricing page lists GPT-5.6 Sol at $4/$20 under promotional pricing that runs until November 30, 2026, Terra at $2/$12, and Luna at $0.20/$1.20. Sol has a 1,050,000-token context and charges more for prompts above 272K input tokens.
- Sonnet 5: $2/$10, 1M context, no long-context surcharge.
- GPT-5.6: per-tier pricing from $0.20/$1.20 (Luna) to $4/$20 (Sol, promotional).
Which to pick
For new projects, start with the successors: Sonnet 5.5 and GPT-6.1 Sol both list at $2/$10 (see Claude Sonnet 5.5 vs GPT-6.1 Sol). For an existing integration, keep Claude Sonnet 5 for agent loops that send long prompts, since its full 1M window costs the same per token. Keep the GPT-5.6 family if you already route between tiers, using Luna for cheap high-volume calls and Sol for the hardest reasoning.
Which should you choose?
Choose Claude Sonnet 5 if...
- →You have a pipeline tuned to Sonnet 5 and want stable behavior while you test Sonnet 5.5
- →You turn thinking off with thinking type disabled or use forced tool_choice, both of which Sonnet 5.5 rejects
- →You need a 1M context window at $2/$10 with the Sonnet 5 request surface
Choose GPT-5.6 if...
- →You have workloads tuned to a GPT-5.6 tier and want stable behavior
- →You want to route requests across Sol, Terra and Luna within one generation