Model Details
Claude Sonnet 3.6
Quick answer: Claude Sonnet 3.6 (API: claude-sonnet-20241022, also referred to as Claude 3.5 Sonnet v2) is Anthropic's October 2024 mid-tier model. It scores 10.4 on the SHADE-Arena sabotage-monitoring benchmark, and was the state-of-the-art instruction-following model at its price tier at launch. At $3/$15 per 1M tokens with a 200K context window.
At a Glance
Where Claude Sonnet 3.6 leads
- 10.4 SHADE-Arena — highest SHADE-Arena score in the Claude 3.x Sonnet family
- State-of-the-art on coding and reasoning at the Sonnet tier at release
- 200K-token context window
- Computer use capability (beta)
- Vision support
Where it lags
- Superseded by Claude Sonnet 4 and Claude 3.7 Sonnet for new deployments
- $3/$15 per 1M — higher cost than Gemini 2.5 Flash at inferior performance on some benchmarks
- April 2024 knowledge cutoff
Best for: Teams running validated production workflows on claude-sonnet-20241022; use cases where the Oct 2024 model version is specifically required.
What Claude Sonnet 3.6 Is
Claude Sonnet 3.6 (also known as Claude 3.5 Sonnet v2) was Anthropic's October 2024 update to the Claude 3.5 Sonnet line. Released October 22, 2024, it was widely benchmarked as the leading coding and reasoning model in its price tier at the time, with notable improvements in agentic tasks and computer use over the June 2024 original.
The model was the first Claude to offer computer use as an API feature (in beta) — allowing it to control a desktop computer to complete tasks. This capability, combined with its SHADE-Arena score, positioned it as a strong choice for agentic workflows.
For new deployments, Claude Sonnet 4 or Claude 3.7 Sonnet provide better overall performance. Claude Sonnet 3.6 remains available for integrations that have been validated specifically on the claude-sonnet-20241022 model version.
Specifications
| Field | Value |
|---|---|
| Organization | Anthropic |
| Parameters | Undisclosed |
| Context window | 200,000 tokens |
| Max output | 8,192 tokens |
| API model ID | claude-3-5-sonnet-20241022 |
| License | Proprietary (API only) |
| Release date | October 22, 2024 |
| Knowledge cutoff | April 2024 |
| Modality | Text + Vision (multimodal) |
| Computer use | Yes (beta) |
Pricing
| Input (per 1M tokens) | Output (per 1M tokens) | |
|---|---|---|
| Anthropic API | $3.00 | $15.00 |
Prompt caching available. Pricing per Anthropic pricing page.
Context Window
Claude Sonnet 3.6 has a 200,000-token context window — roughly 150 pages of text in a single request.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| SHADE-Arena | 10.4 overall success | Anthropic research post | 2025-06 |
Well-known scores from Anthropic launch report (Oct 2024): SWE-bench Verified 49%, HumanEval 93%, MMLU 88.7%, MT-Bench 9.0.
Claude Sonnet 3.6 vs Alternatives
| Model | Context | SHADE-Arena | Price (in/out per 1M) |
|---|---|---|---|
| Claude Sonnet 3.6 | 200K | 10.4 | $3 / $15 |
| Claude Sonnet 3.7 | 200K | 26.2 | $3 / $15 |
| Gemini 2.5 Flash | 1M | — | $0.15 / $0.60 |
| GPT-4.1 | 1M | 5.3 | $2 / $8 |
Claude Sonnet 3.7 scores 26.2 on SHADE-Arena vs 10.4 here — a significant safety improvement at the same price point. For new Sonnet-tier deployments, Claude Sonnet 3.7 or Claude Sonnet 4 are recommended.
Related Models & Benchmarks
- Compare: Claude Sonnet 3.7 · Claude Haiku 3.5 · Gemini 2.5 Flash
- Benchmarks: SHADE-Arena
Frequently Asked Questions
What is Claude Sonnet 3.6?
Claude Sonnet 3.6 (claude-3-5-sonnet-20241022, also called Claude 3.5 Sonnet v2) is Anthropic's October 2024 mid-tier model with 200K context and computer use capability. It has been superseded by Claude Sonnet 3.7 and Claude Sonnet 4 for new deployments.What is the difference between Claude Sonnet 3.5 and 3.6?
Claude Sonnet 3.6 (October 2024) improved on Claude 3.5 Sonnet (June 2024) with better coding performance, higher SWE-bench scores, and enhanced computer use capability. SHADE-Arena: 10.4 for Sonnet 3.6 vs lower for the original 3.5.How much does Claude Sonnet 3.6 cost?
$3.00 per 1M input tokens and $15.00 per 1M output tokens, with prompt caching available.Specs and scores sourced from Anthropic's official announcements (October 2024) and Benchgen evaluations. Pricing cited to the Anthropic pricing page. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.