Anthropic Launches Claude Haiku 5.5: 90% Cheaper, OSWorld Score Jumps to 72.4%
Prompt Engineering · youtube · 2026-10-08
Anthropic has released Claude Haiku 5.5, calling it its cheapest, fastest, and most capable small model yet. Prompts under 100K tokens cost $0.10/M input and $0.50/M output — 90% less than Haiku 4.5 and on par with OpenAI's GPT-6 Luna; above 100K tokens it's half off, with typical workloads estimated 75% cheaper.
On the OSWorld 2.1 computer-use benchmark, Anthropic reports 72.4%, up from 15.7% for Haiku 4.5. It's also the first Haiku with an adjustable effort setting. Anthropic positions it as a workhorse for summaries, classification, and subagent jobs, while recommending Opus 5.5 / Sonnet 5.5 for complex agentic coding. All benchmark numbers are Anthropic's own; independent results aren't in yet.
Also announced: Sonnet 5.5 cache reads are now half price, and Max/Team plans gain monthly API credits ($100 on Max 5x, $200 on Max 20x, up to $500 pooled on Team), rolling out this week. Available on the Claude API, AWS, Google Cloud, and Azure as claude-haiku-5-5.
Related event: Anthropic Launches Claude Haiku 5.5 with ~75% Lower Running Costs(59 posts)→
More from Models
- 'Compression of the RH proof has begun': AI circles buzz over unverified Riemann claim — teortaxesTex · 2026-11-22
- GLM 5.3 and DeepSeek V4.1 Flash resold at 60% off: the economics of open-source giveaways — lxfater · 2026-11-22
- GPT-6.2 Astra and Sol answer more directly but score lower on transparency than Claude — The Decoder · 2026-11-22
- 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