1.0M contextVisionTool callingReasoningPrompt cachingPartner-served

Pricing

USD per 1M tokens
Input$2.00
Output$10.00
Cached input$0.2090% below fresh input
Cache write (5m)$2.50
Cache write (1h)$4.00

Billed per token with no minimum and no monthly fee. Price your workload

Capabilities and limits

Context window
1.0M tokens
Max output
63K tokens
Modality
multimodal
Tool calling
Supported
Vision
Supported
Reasoning
Optional
Reasoning effort
low, medium, high, xhigh, max (default: high)
Prompt caching
Supported (platform_managed)
Cache minimum
4K tokens

Measured performance

7d
Time to first token (p50)
3.0 s
Time to first token (p95)
6.0 s
Throughput (avg)
106.1 tok/s
Throughput (p50)
200.0 tok/s

Measured on real Run BiOS traffic over the trailing 7d. Your figures will vary with prompt shape and region.

Call it now — OpenAI-compatible
curl https://api.runbios.ai/v1/chat/completions \
  -H "Authorization: Bearer $BIOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

Existing OpenAI SDK code works by changing the base URL and the key. API overview

Model card

Curated by Run BiOS · benchmark figures, where present, are the card’s own, not our measurements

Claude Sonnet 5 is the latest model in Anthropic's Sonnet family — the most agentic Sonnet model yet. It can make plans, use tools like browsers and terminals, and run autonomously at a level that, only months earlier, required larger and more expensive models. It is an upgrade to Claude Sonnet 4.6, bringing near-Opus intelligence at Sonnet pricing for coding, agents, and everyday professional work.

Anthropic's launch materials compare Sonnet 5 against Sonnet 4.6 and the more generally capable Opus 4.8: Sonnet 5 is a strict improvement over Sonnet 4.6 across effort levels and offers a much wider range of cost-performance options, matching Opus 4.8's capability on some tasks at higher effort.

Capabilities

  • Effort settings — adjustable reasoning effort trades intelligence for speed and cost per task; Sonnet 5 is substantially more cost-efficient at medium effort than its predecessor. On Run BiOS, Sonnet 5 exposes the low and high reasoning-effort levels.
  • Long context — a 1M-token context window.
  • Multimodal input — text and image input, with text output.
  • Tool use and agents — plans, browses, runs terminals, and works autonomously on multi-step tasks.
  • Prompt caching and batch processing — Anthropic advertises up to 90% cost savings with prompt caching and 50% with batch processing on its own API.

Benchmark results

Results reported in the Claude Sonnet 5 system card and Anthropic's launch materials, alongside the two reference models from Anthropic's own comparison.

Benchmark Claude Sonnet 5 Claude Sonnet 4.6 Claude Opus 4.8
SWE-bench Verified 85.2% 79.6% 88.6%
SWE-bench Pro 63.2% 58.1% 69.2%
Terminal-Bench 2.1 80.4% 67.0% 82.7%
FrontierCode 1.1 Main 38.8%
Humanity's Last Exam (no tools) 43.2% 34.6% 49.8%
Humanity's Last Exam (with tools) 57.4% 46.8% 57.9%
BrowseComp (single agent, max effort) 84.7%
BrowseComp (multi-agent, max effort) 86.6%
OSWorld-Verified (computer use) 81.2% 78.5% 83.4%
GDPval-AA v2 (knowledge work) 1618 Elo 1395 Elo 1615 Elo
HealthBench Professional 57.8% 56.9%
CharXiv Reasoning (with tools) 88.3%

Notes on the numbers, as reported:

  • On Humanity's Last Exam with tools enabled, the gap to Opus 4.8 nearly disappears (57.4% vs 57.9%).
  • On GDPval-AA v2, Sonnet 5 is the first Sonnet model to outscore the concurrent Opus flagship on any benchmark in Anthropic's comparison (1618 vs 1615 Elo).
  • On OSWorld-Verified, Sonnet 5 at its highest effort lands 2.2 points behind Opus 4.8 and well ahead of Sonnet 4.6.
  • Anthropic led the launch with SWE-bench Pro — a deliberately harder, more contamination-resistant variant than the older SWE-bench Verified — where Sonnet 5 improved from 58.1% to 63.2% over Sonnet 4.6.

Safety and evaluations

Anthropic's pre-deployment evaluation of Sonnet 5 is documented in the Claude Sonnet 5 System Card, produced with external experts. Across a broad suite of internal and third-party benchmarks, Anthropic reports clear gains over Claude Sonnet 4.6 in coding, agentic search, multimodal reasoning, and professional-task performance — with the model trailing Anthropic's Opus and Mythos-class models in almost all cases. The system card covers capabilities (coding, agentic search, multimodal reasoning, real-world professional tasks, healthcare, multilingual performance, and life sciences) alongside alignment and safeguards evaluations in depth.