Claude Sonnet 5
Anthropic's most agentic Sonnet yet — near-Opus coding and reasoning at mid-tier pricing.
Overview
What is Claude Sonnet 5
Claude Sonnet 5 is Anthropic's mid-tier large language model released in June 2026. It delivers near-Opus-class agentic performance across coding, reasoning, and tool use at a lower price point, featuring vision, web search, structured output, and adaptive thinking enabled by default.
Running it privately on Venice
On Venice, Claude Sonnet 5 runs under an anonymized privacy tier — your prompts are not stored, profiled, or used for training. You get the same Anthropic model with vision, reasoning, web search, and tool use capabilities, but without the surveillance baggage of Big-Tech platforms. Your interaction history stays zero-retention by default, preserving sovereignty over your data.
Assessment
Strengths and limitations
- Near-Opus performance on coding, reasoning, and agentic tasks at Sonnet-tier pricing.
- Native support for vision, multiple image inputs, tool use, web search, reasoning, and structured JSON output.
- Adaptive thinking is on by default, improving complex reasoning without manual configuration.
- Safer agentic behavior with a lower rate of undesirable behaviors than Claude Sonnet 4.6.
- Code-optimized output well suited for software development and autonomous agents.
- Closed proprietary weights: cannot be self-hosted, fine-tuned, or audited.
- Significantly less capable at cybersecurity tasks than Opus 4.8 and Mythos 5; not optimized for offensive security work.
- Rejects non-default sampling parameters (temperature, top_p, top_k), requiring system-prompt steering instead.
- Not open-source or uncensored.
- Max output limited to 64K tokens on Venice.
Capabilities
What it supports
- Tool use / function calling
- Vision (image input)
- Reasoning
- Web search
- Code-optimized
- Structured output (JSON schema)
- Audio input
- Video input
- Multiple image inputs
- Log probabilities
Specifications
Datasheet
- Maker
- Anthropic
- Released
- June 30, 2026
- API model ID
- claude-sonnet-5
- Modality
- Text, vision, code, agentic tool use
- Open weights
- No — proprietary
- Context window
- 1,000K tokens
- Max output
- 64K tokens
- Capabilities
- Vision, Function calling, Reasoning, Web search, Code-optimized
- Privacy on Venice
- Anonymized — prompts not stored
- Available on Venice since
- Jun 2026
API
Call it from your code
Venice exposes an OpenAI-compatible API. Point your base URL at Venice and pass the model id.
curl https://api.venice.ai/api/v1/chat/completions \
-H "Authorization: Bearer $VENICE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5",
"messages": [{ "role": "user", "content": "Explain quantum tunneling simply." }]
}'Pricing
What it costs on Venice
Billed per token on Venice: $3 per 1M input tokens and $15 per 1M output tokens.
New Venice accounts include a free daily allowance and 500 welcome credits — no credit card required.
Alternatives
How it compares
| Model | Context window | Open weights | Price (Venice) | Best for |
|---|---|---|---|---|
| Claude Sonnet 5 | 1M tokens | No | $3 in · $15 out / 1M | Agentic coding & reasoning |
| Claude Opus 4.8 | 1M tokens | No | $6 in · $30 out / 1M | Maximum capability |
| Claude Sonnet 4.6 | 1M tokens | No | $3.60 in · $18 out / 1M | Prior generation |
| DeepSeek V3.2 | 160K tokens | Yes | $0.33 in · $0.48 out / 1M | Open weights & cost efficiency |
| Kimi K2.6 | 256K tokens | Yes | $0.75 in · $3.50 out / 1M | Open weights & long context |
The best balance of capability and cost in Anthropic's Sonnet family.
Use cases
What it is good for
- 01Autonomous software development and coding agents.
- 02Multi-step knowledge work with tool use, web search, and reasoning.
- 03Vision-enabled workflows such as document analysis and chart interpretation.
- 04High-volume production tasks where speed and cost efficiency matter.
- 05Long-context document summarization and synthesis.
Prompting
Getting better results
Disable adaptive thinking with `thinking: {type: "disabled"}` if you need predictable token counts and faster responses.
Remove `temperature`, `top_p`, and `top_k` from API calls — non-default values return a 400 error.
Steer tone and behavior through detailed system prompts rather than sampling parameters.
Combine tool use with web search for agentic workflows that gather live data and act on it.
Version history
Predecessor with manual extended thinking and configurable sampling parameters.
Current — adaptive thinking by default, improved agentic performance, and new tokenizer.
FAQ
Frequently asked questions
Claude Sonnet 5 is Anthropic's mid-tier large language model released in June 2026. It delivers near-Opus-class agentic performance across coding, reasoning, and tool use at a lower price point, featuring vision, web search, structured output, and adaptive thinking enabled by default.
On Venice, Claude Sonnet 5 costs $3 per 1M input tokens and $15 per 1M output tokens, with cached input priced at $0.30 per 1M tokens. You pay per token with no subscription required.
The model itself is not free, but on Venice you can access it using credits without a subscription. New accounts receive welcome credits to try the model before committing to a paid tier.
No. Claude Sonnet 5 is a proprietary closed-weight model from Anthropic. It cannot be self-hosted, fine-tuned, or audited. If you need open weights, DeepSeek V3.2 or Kimi K2.6 are strong alternatives on Venice.
Yes. On Venice, Claude Sonnet 5 supports tool use and function calling, including structured JSON output, vision input, web search, and reasoning. This makes it well suited for autonomous agents and complex workflows.
Claude Opus 4.8 is Anthropic's flagship and wins on the hardest reasoning and cybersecurity tasks. Claude Sonnet 5 delivers similar agentic performance for coding and everyday knowledge work at half the price, making it the better value for most production workloads.
Yes. Adaptive thinking is on by default, but you can disable it by passing `thinking: {type: "disabled"}` in your API request. Keep in mind that `max_tokens` is a hard limit on total output including thinking tokens.
Yes. Claude Sonnet 5 supports vision and multiple image inputs on Venice, enabling document analysis, chart interpretation, and multimodal agentic workflows alongside its text and code capabilities.
Run Claude Sonnet 5 privately
No prompt logging. No data used for training.