Anthropic releases Claude Sonnet 5.5: Terminal-Bench jumps from 10.3% to 70.6%, tasks up to 30% cheaper
Aiden_Tech_Ai · x · 2026-10-21
Anthropic has released Claude Sonnet 5.5, a model built around code, agentic tasks, and everyday work. The company claims it is 30%+ faster than Sonnet 5 and up to 30% cheaper per task thanks to lower token usage.
Benchmark results:
- 70.6% on Terminal-Bench 4.0 (vs 10.3% for Sonnet 5)
- 55.5% on CursorBench 4.0, about 2 points below Opus 5.5
- Nearly matches Opus 5.5 on GDPval-AA
- Better at long tasks, images, documents, spreadsheets, and presentations
The biggest jump is in coding: Anthropic says Sonnet 5.5 moves through codebases faster with fewer tool calls and fewer steps per task.
API pricing: $2/1M input tokens, $10/1M output tokens, $0.20/1M cache reads. API ID is claude-sonnet-5-5; already available in Claude, the API, AWS, Google Cloud, and Azure.
More from coding & agent
- Tamper-proof agent logs still can't prove the agent did the job: who reads the record back? — ZealousidealAct9661 · 2026-10-21
- One prompt: Claude builds an animated mech spider in Blender via MCP, first try — sidahuj · 2026-10-21
- Open-source ComfyUI film studio: four director agents auto-generate movies with frame retouching — Goat_bless · 2026-10-21
- OpenAI Details Coordinated Disclosure of Critical Vulnerabilities Found by Agents — deliciousmark12 · 2026-10-21
- Dev: coding agents make every idea 10 minutes away — and that's the problem — feregri_no · 2026-10-21
- Chops: open-source macOS app organizes coding agent skills across Claude Code, Cursor, Codex and more — tom_doerr · 2026-10-21