Anthropic launches Claude Haiku 5.5: ~75% cheaper, 72.4% on OSWorld, Sonnet cache halved
claudeai · x · 2026-10-08
Anthropic released Claude Haiku 5.5, its cheapest/fastest small model, averaging 75% cheaper than Haiku 4.5.
Key points
- Major gains in coding, computer use, and knowledge work; works as a subagent with Opus 5.5 / Sonnet 5.5; fast enough for live support and browser use.
- Benchmarks: 72.4% on OSWorld 2.1 offline subset (vs 15.7% for Haiku 4.5); HLE 45.9% no-tools / 57.4% with tools; Terminal-Bench 4.0 39.2%; FrontierCode 1.1 Main 46.4%.
- First Haiku with an adjustable effort setting; best value under 100K tokens (90% of prior Haiku traffic).
- Sonnet 5.5 cache reads halved to $0.10/M tokens (20% cheaper long-running work); new monthly API credits for Max/Team subscribers.
- Live on AWS, Google Cloud, Azure; big improvements on alignment evals with far fewer misaligned behaviors.
Related event: Anthropic Launches Claude Haiku 5.5 with ~75% Lower Running Costs(59 posts)→
More from Models
- Baseline model test: no persona, no forced formatting, no safety bypasses observed — Sauers_ · 2026-11-22
- 2-day Gemini stress test: $4.99 Google AI Pro nearly maxed, orchestrator UI revealed — lipeng0820 · 2026-11-22
- Timed eval task asks users to deliberately make Claude fail at real work in 12 minutes — Icy-Measurement10 · 2026-11-22
- Tesla FSD 15's real-time view isn't 3D rendering — it's the AI model's own synthetic perception — mark_k · 2026-11-22
- Gemini Black Friday half-price: JP-region trick gets Pro annual plan for $91.54 — huangyun_122 · 2026-11-22
- Microsoft's USB4 RDMA lets mini PCs pool memory to run Qwen3-235B — JFPuget · 2026-11-22