CLAUDE SONNET 5
Released June 30, 2026. 1M context, adaptive thinking, vision, native web search. Intro pricing $2/$10 per million tokens through Aug 31, 2026 — then $3/$15 standard.
Claude Sonnet 5 is Anthropic's mid-tier agentic model, released June 30, 2026, and the current default answer to 'which Claude should I use?'. Its pitch is simple: close-to-Opus-4.8 capability at a third of the cost. Standard pricing is $3 per million input tokens and $15 per million output — the same sticker as Sonnet 4.6 — but Anthropic is running introductory pricing of $2/$10 through August 31, 2026, making it temporarily one of the best-value frontier models on the market. It carries the full mid-2026 Claude feature set: 1M-token context, vision input, native web search, and adaptive thinking. Versus its predecessor Sonnet 4.6, the capability jump is generational — Sonnet 5 was trained as an agentic model first, so tool-calling runs and multi-file coding tasks land closer to Opus quality than any previous Sonnet. The case for Sonnet 5: daily coding, agents, and general work — the bulk of real usage. The case against: the last 10–15% of refactor quality and prose voice (Opus 4.8), and the hardest reasoning (Fable 5).
Sonnet 4.6 remains in the lineup at $3/$15 and stays a strong pick, but Sonnet 5 is a generational step, not a point release: it was trained as an agentic model first, so multi-step tool runs and multi-file code edits land materially closer to Opus quality. Both ship 1M context, vision, and native web search. Until August 31, 2026 the decision is easy — Sonnet 5 is both better and cheaper ($2/$10 intro vs $3/$15). From September 1 they cost the same, and Sonnet 5 is simply the stronger model at that sticker.
Sonnet 5 is available on every paid Council AI plan (Plus, Pro, Ultra) and its cost profile makes it a natural everyday council seat: strong enough to anchor quality, cheap enough to include in nearly every fan-out — especially at intro pricing. Councils run it in parallel with models from up to 8 other labs, and the moderator's agreement score tells you when Sonnet 5's answer stands with the consensus or against it. A popular budget-conscious lineup is Sonnet 5 + GPT-5.6 Terra + Gemini 3.5 Flash: three labs, frontier-adjacent quality, mid-tier spend.
Intro pricing is $2.00 per million input tokens and $10.00 per million output through August 31, 2026. From September 1 it moves to standard $3.00/$15.00 — the same sticker as Sonnet 4.6.
1,000,000 tokens, with vision input and native web search included.
Close, deliberately: Anthropic positions it as close-to-Opus-4.8 capability at a third of the cost. Opus 4.8 keeps an edge on the hardest refactors and voice-sensitive writing; for most daily coding and agent work the gap is hard to notice.
Yes. Through Aug 31, 2026 Sonnet 5 is both better and cheaper (intro $2/$10 vs 4.6's $3/$15). After that they cost the same and Sonnet 5 is simply the stronger model, especially on agentic tasks.
Yes — on all paid plans (Plus, Pro, Ultra). It runs in parallel councils with models from 8 other labs, and the moderator scores cross-model agreement on every answer.