SHELFCRITTER AI Digest

MODEL

Claude Sonnet 5.5

modeltopic-noteanthropic

Overview

Claude Sonnet 5.5 is Anthropic‘s Sept 28, 2026 successor to Claude Sonnet 5 — a Sonnet-tier release that holds the $2/$10 per Mtok list price with $0.20/Mtok cache reads and a 1M-token context window, while collapsing the Opus-vs-Sonnet gap on coding and agentic-workflow axes. Terminal-Bench 4.0 jumps 10.3% (Sonnet 5) → 70.6% and GDPval-AA v2.1 lands within two Elo of Claude Opus 5.5. Free-tier claude.ai defaults to Sonnet 5.5 immediately on release, and Claude Code v2.1.284 wires claude-sonnet-5-5 as the default Sonnet in the CLI the same day.

Timeline

  • 2026-09-29-AI-Digest — Anthropic released Claude Sonnet 5.5 on 2026-09-28 at unchanged $2/$10 per Mtok list price with $0.20/Mtok cache reads and a 1M-token context window; Anthropic marketing frames it as ~30% faster and up to ~30% cheaper per task. Independently, Terminal-Bench 4.0 jumps from 10.3% (Sonnet 5) to 70.6% — a 6.8× leap on the same tier — and GDPval-AA v2.1 puts Sonnet 5.5 two Elo points shy of Claude Opus 5.5. Free-tier claude.ai now defaults to Sonnet 5.5 (previously Claude Sonnet 5), and same-day Claude Code v2.1.284 wires claude-sonnet-5-5 as the default Sonnet in the CLI. Load-bearing softener: the “30% faster / 30% cheaper” line is Anthropic marketing copy — read it as per-task token efficiency at unchanged headline pricing, not a list-price cut. What is genuinely new is the Terminal-Bench delta and the free-tier distribution shift. Reframe worth carrying: Sonnet 5.5 collapses Opus-vs-Sonnet on coding/agentic axes at Sonnet pricing, and free-tier claude.ai now runs on the frontier-Sonnet tier, not Anthropic dropped Sonnet prices. Log against MOC - Major Companies and MOC - Developer Tools.

Key Developments

  1. Terminal-Bench 4.0 Jumps 10.3% → 70.6% on the Same Sonnet Tier (September 28, 2026): The independent Terminal-Bench 4.0 delta is the load-bearing capability datum — a 6.8× jump inside the Sonnet tier, not a marketing composite. Pairs with GDPval-AA v2.1 landing two Elo shy of Claude Opus 5.5 to define the “Sonnet-tier collapses onto Opus-tier capability on coding + agentic axes” reframe. The corpus should stop treating Opus as the necessary default for tool-use-heavy agent stacks — Sonnet 5.5 at unchanged $2/$10 reshapes that calculus. Covered in 2026-09-29-AI-Digest.

  2. Free-Tier claude.ai Now Defaults to Sonnet 5.5 (September 28, 2026): The distribution shift — free-tier claude.ai deploying on the frontier-Sonnet tier the same day Sonnet 5.5 released — is what turns the capability delta into a practitioner-visible change inside the week. Load-bearing framing: frontier capability at cheaper-tier pricing + free-tier distribution, not Anthropic dropped prices. Covered in 2026-09-29-AI-Digest.

  3. Same-Day Claude Code Substrate Cutover in v2.1.284 (September 28, 2026): Claude Code v2.1.284 swapped the Sonnet default over to claude-sonnet-5-5 on the same UTC window Sonnet 5.5 shipped — continuing the model-and-substrate same-day cadence the Claude Opus 5.5 / v2.1.280 launch already established, and resolving the “compound-quiet” watch item flagged in 2026-09-27-AI-Digest and 2026-09-28-AI-Digest within ~24 hours. Covered in 2026-09-29-AI-Digest.

  4. Unchanged $2/$10 List Price With $0.20/Mtok Cache Reads (September 28, 2026): List pricing is held from Sonnet 5’s post-Aug 11 durable rate (2026-08-12-AI-Digest) — the corpus’s manager-delegates-to-cheaper-worker Advisor / Orchestrator economics survive the release without recalibration. Anthropic’s ~30% faster / cheaper claim is marketing copy — read as per-task token efficiency at unchanged headline pricing, not a list-price cut. Covered in 2026-09-29-AI-Digest.

  • 2026-10-01-AI-Digest — Claude Sonnet 5.5 ($2/$10, Sep 28) is now one of four frontier labs on the convergent mid-tier line as Google ships Gemini 4 Argon at $2/$10 introductory the same per-token line. Four frontier labs inside a single quarter — Grok 4.7 ($2/$6, Sep 21), Sonnet 5.5 ($2/$10, Sep 28), GPT-6.1 Sol ($2/$10, Sep 29), and Gemini 4 Argon ($2/$10 intro, Oct 1). Framing to carry: convergent competitive repricing, not coordination — Argon’s post-introductory $4/$20 standard rate is the first visible signal the floor may not hold through 2027; Sonnet 5.5’s price is listed as the durable $2/$10 from the Sept 28 release with no scheduled step-up, so the Sonnet-tier is the anchor the other three labs are now benchmarked against.

See also: Anthropic, Claude Sonnet 5, Claude Opus 5.5, Claude Code, Gemini 4 Argon, GPT-6.1 Sol, Grok 4.7, MOC - Major Companies, MOC - Developer Tools, MOC - Agentic Coding.