Back to Gap Map Model components / Fine-tuned / chat models

Claude Haiku

closed / Overall score: 4.0(strong)

Claude Haiku is Anthropic's small, fast tier, built for high-volume, low-latency coding and chat where cost and speed matter more than peak reasoning. It sits beneath Sonnet and Opus, and the name persists across successive generations rather than pinning to one release. Weights are closed, and it is available through the Anthropic API and the Claude apps.

Tier entry covering successive Haiku generations; Haiku 4.5 is the release scored.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only)

Claude Haiku is closed by construction across every generation: no weights, training data or code are released. Haiku 4.5, the current release, is sold only through Anthropic's own API and app surfaces, with no weight download offered.

https://www.anthropic.com/claude/haiku recorded 2026-09-22
"Pricing for Haiku 4.5 on the Claude Platform starts at $1 per million input tokens and $5 per million output tokens"; distributed via Claude.ai, the Claude Platform, Amazon Bedrock, Vertex AI and Microsoft Foundry; no weights, corpus or training code offered

Adoption

4 medium confidence

4.0

Haiku is positioned for high-volume, low-latency use, but Anthropic publishes no per-model usage figure, so this reading is directional.

https://www.anthropic.com/claude/haiku recorded 2026-08-13
"advanced capabilities at a price that makes Claude practical for scaled deployments, from powering free tier products to budget-conscious applications"; latency-sensitive, coding sub-agent and research sub-agent use cases; no usage or user figure published

Capability

4 medium confidence

4.0

Haiku 4.5 matches the prior Sonnet 4's performance on coding, computer use and agent tasks, and Anthropic reports 73.3% on SWE-bench Verified for it, but it remains below the current Sonnet, Opus and Fable tiers. The SWE-bench Verified leaderboard has taken no submissions since 2025-12-15, so that figure is vendor-reported rather than independently confirmed.

https://www.anthropic.com/claude/haiku recorded 2026-09-29
Product page: "matching Sonnet 4’s performance on coding, computer use, and agent tasks. Claude Haiku 4.5 scores 73.3% on SWE-bench Verified".
https://www.anthropic.com/news/claude-haiku-4-5 recorded 2026-09-29
Launch post: "Claude Haiku 4.5 even surpasses Claude Sonnet 4 at certain tasks, like using computers"; the SWE-bench Verified score is vendor-reported.

Verified 2026-09-22, except some scores, still awaiting re-confirmation