This page requires JavaScript to display.

Claude Opus 5.5 launch

AI & tech · Heat 2.3 / 10 · 3 shows · 25m airtime

Anthropic announced Claude Opus 5.5 on September 23, 2026, claiming Fable 5.1-level performance at 40% lower cost. Reporting has since established pricing at $4 per million input and $20 per million output tokens, output roughly 30% faster than Opus 5, a 66.4% score on Terminal Bench 4.0 against Fable 5.1's 55.8%, and a verified 285 tokens per second at max with 50% more tokens per dollar than OpenAI's fast mode. Every calls it the most readable prose yet from an Anthropic or OpenAI model. Anthropic now requires thinking blocks preserved across turns and shipped an API compaction feature, and Opus 5.5 does not carry Fable's data retention requirements. What remains contested is how much the benchmark lead translates into practice.

How 3 podcasts covered it

  1. The AI Daily Brief, 2026-09-23 (3:17 to 4:25)
    Opens with Opus 5.5's Terminal Bench score
    Opens the episode with Anthropic's Opus 5.5 announcement: Fable 5.1-level performance at 40% lower cost, 66.4% on Terminal Bench 4.0 against Fable 5.1's 55.8%, and gains on Cursor Bench, Frontier Code v1.1 and Humanity's Last Exam.
    New angle: Opus 5.5 scores 66.4% on Terminal Bench 4.0 versus Fable 5.1's 55.8%, at 40% lower cost.
  2. The AI Daily Brief, 2026-09-23 (4:25 to 5:32)
    Reads the Opus 5.5 pricing sheet
    Reads the Opus 5.5 pricing sheet: $4 per million input and $20 per million output tokens, with further efficiencies bringing the real gain to about 40%, plus 30% faster output and higher 5-hour limits on Pro, Max and Team plans.
    New angle: Opus 5.5 is priced at $4 per million input and $20 per million output tokens, with output about 30% faster than Opus 5.
  3. The AI Daily Brief, 2026-09-23 (9:51 to 10:43)
    Adds the Intelligence Index top score of 58
    Collects first reactions to Opus 5.5 and adds a benchmark number: the Intelligence Index top score moves to 58, five points above Fable 5.1 and GPT-6 Astra at 53, with a 20% price cut.
  4. The AI Daily Brief, 2026-09-23 (14:22 to 16:38)
    Collects the viral 3D and visual demos
    Collects the visual and 3D demos that dominated social media reaction to Opus 5.5 - Blender claymations, code-generated Golden Gate Bridge video, a 1906 Market Street scene, and Python-generated paintings - and notes the model appears to outpace GPT-6 Astra on spatial and 3D work.
    New angle: Jake Eaton's viral paintings are ~7,500 lines of Python generated pixel by pixel, with no image model or reference images.
  5. The AI Daily Brief, 2026-09-23 (17:20 to 19:21)
    Picks up the personality and writing gains
    Picks up the personality and writing side of the Opus 5.5 launch: Anthropic's own account says it addresses Opus 5 feedback and follows writing rules, and Every calls it the most readable prose from an Anthropic or OpenAI model.
    New angle: Every calls Opus 5.5 the most readable prose it has seen from an Anthropic or OpenAI model.
  6. Equity, 2026-09-25 (5:45 to 7:01)
    Frames Opus 5.5 as an enterprise play
    Notes that Opus 5.5 is Anthropic's most expensive tier and frames it as an enterprise play, part of a broader argument about both labs pushing into enterprise applications. Reaction and framing, not a new development.
  7. The AI Daily Brief, 2026-10-13 (0:54 to 2:05)
    Leads with verified inference speed
    Leads the headlines with inference speed: Anthropic now serves Opus 5.5 at a verified 285 tokens per second at max, against a theoretical 280 for OpenAI's fast mode, and can serve 50% more tokens per dollar. Notes fast mode is not included in regular subscriptions.
    New angle: Anthropic serves Opus 5.5 at a verified 285 tokens per second at max, and 50% more tokens per dollar than OpenAI's fast mode.
  8. The AI Daily Brief, 2026-10-20 (22:31 to 23:40)
    Recaps the backlash before the 5.5 launch
    Recaps the Opus 5 backlash that preceded the 5.5 launch - Claire Vo's video, Reddit complaints about a passive-aggressive model, and the team member's apology. Background for the launch, with nothing new on the record.
  9. Latent Space, 2026-10-27 (7:31 to 14:26)
    Explains the anti-distillation work shipped with Opus 5.5
    Anthropic engineers describe the anti-distillation work shipped alongside Opus 5.5: required thinking-block preservation across turns, a new API compaction feature, and outreach to open source harness maintainers and large customers. They say distillation is now harder by the measures they can observe, and frame it as a safety concern rather than a commercial one.
    New angle: Anthropic now requires thinking blocks preserved across turns and shipped an API compaction feature to replace broken customer workarounds.
  10. Latent Space, 2026-10-27 (17:48 to 23:13)
    Notes Opus 5.5 drops Fable's retention rules
    Jiang walks through what a launch now involves: per-model data retention policy, cyber and life science verification programs, and an authorization system that threads model and program access through the stack. She puts on the record that Opus 5.5 does not carry Fable's data retention requirements.
    New angle: Opus 5.5 does not carry the data retention requirements that Fable has.
  11. Latent Space, 2026-10-27 (1:10:56 to 1:12:28)
    Explains the pacing contradiction on Opus 5.5
    Jiang answers the pacing contradiction: Opus 5.5 shipped eleven days after Anthropic called for pacing because it cleared safety checks across infrastructure security, online safeguards and offline classification. She concedes safeguard false positives remain a frustrating user experience. Framing and explanation, no new development.
Unpacking...