Claude Fable 5
The newest Anthropic model — first Mythos-class made generally available.
Reference benchmarks for Claude Fable 5
These are external benchmarks I pulled from the source comparison guides on agentos.guide — SWE-bench Verified, DRACO, Kilo plan rubric, build-time measurements, vendor-reported coding scores. They are not goldiebench medal scores (those come only from same-prompt one-shot creative coding tasks in the matrix). I surface them here so the spec sheet for Claude Fable 5 is honest about what's measured.
What is Claude Fable 5?
Claude Fable 5 is the Anthropic frontier model with a 200,000 tokens (1M with extended thinking) context window, released 2026-06-09. Tagline: The newest Anthropic model — first Mythos-class made generally available.. Official source: anthropic.com/claude.
Pricing detail. Released alongside Mythos 5 on June 9, 2026 as the publicly-available member of the new Mythos class. Premium per-token pricing on the Anthropic API; available everywhere Opus 4.8 ships.
How I use it inside the Agent OS. Selected from Agent OS for the highest-stakes work — it replaced Opus 4.8 as the safety net on hard prompts. Its four core 3D games were rebuilt to showcase quality with the threejs-game-director skill, lifting the full 42-task bench to 8.14 avg — the #1 solo model, behind only the Fusion and MoA ensembles.
What I built with Claude Fable 5
Every model on Goldie Bench gets the same fixed prompt set — one shot, single HTML file out — and I score the result 0–10 inside the Agent Operating System. Here's what Claude Fable 5 shipped on the bench: 47 one-shot demos across 200,000 tokens (1M with extended thinking) of context. Of those, 47 are scored against the field with my honest 0–10 from the source guides at agentos.guide.
Strengths
- Now the top SOLO model on this bench — 8.14 avg, #3 overall, edging Grok (8.13); only the Fusion (8.60) and Hermes MoA (8.38) ensembles rank higher
- 15 medals across 42 tasks (5 gold, 2 silver, 8 bronze) — shader/GPU physics is its superpower (Cornell-box path tracer 8.7, black-hole lensing 8.7, synthwave outrun 8.7)
- Its four core 3D games (crypt, skyrim, twilightvale, voxelcraft) rebuilt to showcase quality with the threejs-game-director skill — authored heroes, layered worlds, PBR materials, cohesive HUDs, all 8.8–9.0
- Beats Opus 4.8 head-to-head on the majority of tasks; tops external SWE-bench Verified at 95.0% in Julian's three-dragons writeup
Trade-offs
- Its hardest one-shots (crypt, twilightvale) black-screened on three.js r128 API drift — the scored builds are agentic rebuilds, not the raw first pass, and crypt's AAA rebuild needed a one-line emissive patch
- The showcase ceiling shown here needs the threejs-game-director scaffolding baked into the prompt — a bare one-shot lands lower (7.72 avg)
- Premium $10/$50 per-M pricing — you're paying for reasoning depth; cheaper models stay competitive on pure one-shot visuals
Best for
- Mission-critical one-shot builds where you want Anthropic's newest reasoning
- Long-context work using extended thinking up to 1M tokens
- Plan-heavy multi-step tasks where intelligence in the plan matters more than the build
Every benchmark — Claude Fable 5's full scorecard
All 47 scored tasks, best first — the judge's 0–10 on the same rubric as the whole field. Click any bar for that task's cross-model page, or open this scorecard in the interactive graphs. Full editorial breakdown with judge quotes and sourced outside research: the Claude Fable 5 deep dive →.
Every demo by Claude Fable 5
47 live demos, sorted by category. Click any tile to play the actual one-shot result. Verdicts and 0–10 scores are pulled from the source guides where I posted them publicly.
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVE
▶ LIVECompare Claude Fable 5 against every other model
Every head-to-head featuring Claude Fable 5. Verdicts shown for scored pairs.
See all 66 comparisons across every model →
Quick pill index
Direct comparisons against every other scored model on the bench:
Claude Fable 5 vs Fusion Claude Fable 5 vs Claude Opus 5 Claude Fable 5 vs Hermes MoA Claude Fable 5 vs GPT-5.6 Sol Claude Fable 5 vs Qwen 3.8 Claude Fable 5 vs Grok Claude Fable 5 vs MiniMax M3 Claude Fable 5 vs Fugu Ultra Claude Fable 5 vs Kimi K3 Claude Fable 5 vs GLM-5.2 Claude Fable 5 vs Fugu Mini Claude Fable 5 vs Opus 4.8 Claude Fable 5 vs Kimi K2.7 Claude Fable 5 vs Qwable 5 27B Coder Claude Fable 5 vs Gemini 3.6 Flash Claude Fable 5 vs Claude Sonnet 5 Claude Fable 5 vs Qwen 3.7 Claude Fable 5 vs Fugu Ultra 1.1 Claude Fable 5 vs Inkling Claude Fable 5 vs Agents-A1 Claude Fable 5 vs Gemma 4 12B · MLX Claude Fable 5 vs Laguna XS 2.1 Claude Fable 5 vs Qwythos 9B Claude Fable 5 vs LongCat-2.0 Claude Fable 5 vs Hy3 Claude Fable 5 vs Gemma-4 12B CoderRead more on agentos.guide: /claude-fable-5, /claude-fable-5-use-cases, /claude-5-agent-os, /claude-5-ai-seo, /claude-hermes-olympus-protocol
Claude Fable 5 — frequently asked
What is Claude Fable 5?
Claude Fable 5 is Anthropic's AI model — The newest Anthropic model — first Mythos-class made generally available. It has a 200K tokens context window and was released 2026-06-09.
How good is Claude Fable 5 at coding and one-shot builds?
On the GoldieBench one-shot build benchmark it averages 8.10/10 across 47 scored tasks, with 4 gold, 2 silver and 1 bronze medals.
How much does Claude Fable 5 cost?
$10 / $50 per M tokens. Released alongside Mythos 5 on June 9, 2026 as the publicly-available member of the new Mythos class. Premium per-token pricing on the Anthropic API; available everywhere Opus 4.8 ships.
Where can I see Claude Fable 5 demos?
Every one-shot build is live and playable on this page and on the GoldieBench compare matrix — same prompt as every other model, no retries.
Run this stack yourself.
Every demo on this bench was built inside the Agent Operating System — one prompt, one shot, single HTML file out. The Agent OS, the prompts, the templates, the weekly walkthroughs and 4,000+ founders shipping with it every day all live inside the AI Profit Boardroom.