Claude Sonnet 5
Language AvailableCompare models
ComparingClaude Sonnet 5with
Pick a model above to see the comparison.
anthropic/claude-sonnet-5 · by Anthropic · decoder-only-transformer
Pricing — 1 offering(s)
Input tokens
- $2.00 / 1M tokens (input) Current 2026-07-01 → present
- $3.00 / 1M tokens (input) Future 2026-09-01
Output tokens
- $10.00 / 1M tokens (output) Current 2026-07-01 → present
- $15.00 / 1M tokens (output) Future 2026-09-01
Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.
Capability profile
reasoning strong
coding strong
tool use strong
instruction following strong
context window strong
multilingual strong
speed moderate
cost efficiency moderate
Coding benchmarks
View full leaderboard →| Benchmark | Score | Harness | Source |
|---|---|---|---|
| swe-bench-verified ↗ | 85.2% 2026-06 | agentless 2.0 | 🏢 vendor ↗ |
| swe-bench-pro | 63.2% 2026-06 | agentless | 🏢 vendor ↗ |
| terminal-bench-2-1 | 80.4% 2026-06 | terminus-2 | 🏢 vendor ↗ |
Operator guidance
The new default for most Claude use cases — strong across reasoning, coding, and tool use at Sonnet-tier pricing. Step up to Opus 4.8 for the most demanding agentic or reasoning tasks; drop to Haiku 4.5 for cost/latency-sensitive work. Note the introductory pricing window (through August 31 2026).
Use cases
- Default model for Claude Free and Pro plans (July 2026)
- Agentic coding, complex reasoning, and knowledge work at Sonnet pricing
- Near-Opus 4.8 quality for most tasks at a lower cost
Limitations
- Not as strong as Opus 4.8 on the hardest agentic/reasoning tasks
- Updated tokenizer produces ~30% more tokens than Sonnet 4.6 for the same input — factor this into cost-per-task comparisons vs. earlier models
- Standard $3/$15 pricing resumes September 1 2026 (introductory $2/$10 through August 31 2026)
- Capability ratings are vendor-reported at launch; independent benchmark corroboration pending