
Claude Sonnet 5
Anthropic's June 2026 Sonnet model for agentic work, priced at $2/$10 per million tokens, now a legacy model.
Claude Sonnet 5 is Anthropic's Sonnet model released June 30, 2026, built to plan, use tools such as browsers and terminals, and run autonomously, with performance Anthropic puts close to Opus 4.8. On the Claude API it costs $2 per million input tokens and $10 per million output tokens, a price made permanent in August 2026.
About Claude Sonnet 5
The launch post calls Sonnet 5 a substantial improvement over Sonnet 4.6 in reasoning, tool use, coding and knowledge work. You choose an effort level to trade cost against performance: Anthropic reports substantially better cost efficiency at medium effort, and at higher effort Sonnet 5 can match Opus 4.8 on some tasks, shown in charts for BrowseComp and OSWorld-Verified. The full assessment is in the Claude Sonnet 5 System Card.
Anthropic's safety evaluations found fewer undesirable behaviors than in Sonnet 4.6 and better refusal of malicious requests and resistance to prompt-injection hijacking, though somewhat more misaligned behavior than Opus 4.8 and Claude Mythos Preview. Sonnet 5 was not deliberately trained on cybersecurity tasks, never developed a full working exploit in a Firefox exploit evaluation, and launched with real-time cyber safeguards on by default, the same as in Claude Opus 4.7 and 4.8.
At launch, Sonnet 5 became the default model for the Free and Pro plans and was available to Max, Team and Enterprise users, and developers call it as claude-sonnet-5 through the Claude API. On August 10, 2026 Anthropic made the $2/$10 launch price permanent, dropping a planned move to $3/$15 on September 1. The pricing page now lists Sonnet 5 among legacy models, with cache reads at $0.20 and writes at $2.50 per million tokens.
Sonnet 5 uses an updated tokenizer, similar to the change in Claude Opus 4.7, so the same input can map to roughly 1.0 to 1.35 times as many tokens depending on content. Anthropic raised rate limits across Chat, Cowork, Claude Code and the Claude Platform to cover the higher token use of higher effort levels.
Sources: anthropic.com/news/claude-sonnet-5, claude.com/pricing and platform.claude.com/docs/en/home, read 2026-10-26
Claude Sonnet 5 pros and cons
Pros
- Anthropic puts its performance close to Opus 4.8 at less than half the per-token price: $2 and $10 per million input and output tokens against $5 and $25.
- Lower rates of hallucination and sycophancy than Sonnet 4.6, according to Anthropic's safety evaluations.
- Cyber safeguards that detect and block dangerous cyber usage in real time are on by default.
- The pricing page offers 50% off with batch processing.
Cons
- The pricing page now files it under legacy models, and Sonnet 5.5, the newer Sonnet, has the same $2 input and $10 output price.
- Its updated tokenizer can count the same input as up to 1.35 times as many tokens.
- For cybersecurity work that needs reduced guardrails, Anthropic recommends Claude Opus 4.8 instead.
Tagged
- Supports Claude
- Web-based
- Agentic Mode
Claude Sonnet 5 alternatives
The closest alternatives to Claude Sonnet 5.
- 01Paid →Claude Opus 5
Anthropic's Opus model from Jul 2026 for coding and agents, at $5/$25 per million tokens, now a legacy model.
- 02→Kimi K3
Moonshot AI's 2.8T-parameter model with native vision and a 1M-token context for long-horizon coding.
- 03→DeepSeek-V4-Flash-0731
Official release of DeepSeek-V4-Flash, with stronger agentic skills and open weights under the MIT License.
Related reading
Featured on Vibedonalds
Own Claude Sonnet 5? Add this badge to your site to show you’re listed — and link back to your profile here.
<a href="https://vibedonalds.com/tools/claude-sonnet-5" target="_blank" rel="noopener nofollow">
<img src="https://vibedonalds.com/badge/featured-on-vibedonalds.svg" alt="Claude Sonnet 5 — Featured on Vibedonalds" width="240" height="60" loading="lazy" />
</a>Frequently asked questions
- What is Claude Sonnet 5?
- Claude Sonnet 5 is Anthropic's June 2026 Sonnet model for agentic work, priced at $2/$10 per million tokens, now a legacy model.
- What are the alternatives to Claude Sonnet 5?
- The closest alternatives to Claude Sonnet 5 are Claude Opus 5, Kimi K3, DeepSeek-V4-Flash-0731.
- What are the downsides of Claude Sonnet 5?
- The pricing page now files it under legacy models, and Sonnet 5.5, the newer Sonnet, has the same $2 input and $10 output price. Its updated tokenizer can count the same input as up to 1.35 times as many tokens. For cybersecurity work that needs reduced guardrails, Anthropic recommends Claude Opus 4.8 instead.
- How much does Claude Sonnet 5 cost?
- Through the Claude API, $2 per million input tokens and $10 per million output tokens. Anthropic made that launch price permanent on August 10, 2026, dropping a planned rise to $3 and $15 on September 1, and the pricing page lists cache reads at $0.20 and cache writes at $2.50 per million tokens.
- When was Claude Sonnet 5 released?
- Anthropic announced it on June 30, 2026, and made it available that day across all Claude plans, as the default model for Free and Pro, and to developers through the Claude API as claude-sonnet-5.
- Claude Sonnet 5 vs Opus 4.8: which should I use?
- Anthropic says Sonnet 5 performs close to Opus 4.8 and can match it on some tasks at higher effort, for $2/$10 per million tokens against $5/$25. Opus 4.8 showed less misaligned behavior and stronger cyber skills, and Anthropic recommends it for cybersecurity work that needs reduced guardrails.
- How is Claude Sonnet 5 different from Sonnet 4.6?
- Anthropic calls it a substantial improvement in reasoning, tool use, coding and knowledge work, with lower rates of hallucination and sycophancy. It costs $2/$10 per million tokens against Sonnet 4.6's $3/$15, but its new tokenizer can map the same input to roughly 1.0 to 1.35 times as many tokens.
- What does xhigh effort mean for Claude Sonnet 5?
- xhigh is the extra high effort level. Higher effort levels use more tokens, and Anthropic raised rate limits across Chat, Cowork, Claude Code and the Claude Platform to accommodate them; it reports substantially improved cost efficiency at medium effort and Opus 4.8-level results on some tasks at higher effort.