Skip to main content
AIJun 10, 20267 min

What Is Claude Fable 5? Anthropic's New Mythos-Class Model (Free Until June 22, 2026)

Claude Fable 5 is Anthropic's new model launched June 9, 2026 — a **Mythos-class** model Anthropic calls "more capable than any we've made generally available." SOTA on nearly every benchmark, game-changing for software engineering, vision, and long-context work — with a safety fallback to Opus 4.8 in <5% of sessions. API pricing $10/$50 per 1M tokens. Free for Pro/Max/Team/Enterprise subscribers through June 22. Full breakdown + when Thai users should use it.

Claude Fable 5 Anthropic Mythos-class AI model 2026 - CherCode

Quick answer: Claude Fable 5 is Anthropic's new model launched June 9, 2026 — a Mythos-class model (a tier above Opus) Anthropic calls "more capable than any we've made generally available." SOTA on nearly every benchmark — strong on coding, vision, long-context, finance, analytics. Key catch: a safety fallback to Opus 4.8 kicks in for cybersecurity/biology/distillation queries (<5% of sessions). API pricing $10 input / $50 output per 1M tokens. Free for Pro/Max/Team/Enterprise subscribers through June 22, then credits-based.

📌 Big picture: this is the first time Anthropic releases a Mythos-class model to the public — previously limited to Project Glasswing (cyberdefense partners) because it can find critical-severity vulnerabilities in OSes and browsers. Fable 5 = the same model, plus guardrails.

Fable 5 vs Mythos 5 — What's the Difference

The underlying model is identical. Difference is in guardrails: Mythos 5 = full capability, limited to cyberdefense partners / approved biology researchers (with cyber guardrails). Fable 5 = same model + full guardrails, for general use — any dev or business paying for Claude can now access Mythos-class. Anthropic released a Mythos Preview in April 2026 — that release found thousands of critical/severe vulnerabilities in major OSes and browsers, which is why broader rollout was held. Fable 5 is how Anthropic ships it safely to the public.

Confirmed Capabilities + Benchmarks

Official numbers from Anthropic (June 9, 2026):

Benchmark / TaskFable 5 result
Analytics core benchmark>90% (first model to clear it)
Slay the Spire (long-context + memory) Opus 4.8 performance
Pokémon FireRed (vision-only)Cleared — prior versions needed helper tools
FrontierCodeHighest score among frontier models (medium effort)
Hebbia Finance BenchmarkHighest score + IMC confirms strong trading analysis
Stripe case studyRuby codebase migration: 2 months → 1 day

💡 Key signal: Anthropic states "the longer and more complex the task, the larger Fable 5's lead" = ideal for agentic workflows + long-document analysis, not short chats.

Claude Fable 5 benchmarks: Analytics 90%, Long-context 3x Opus, FrontierCode top score

Pricing + Free Window (Critical for Thai users)

Anthropic splits access into 2 channels:

ChannelCost / ConditionStart / End
Pro / Max / Team / Enterprise subscriberFree (included)June 9-22, 2026
Pro / Max / Team / Enterprise after June 22Usage credits (paid add-on)June 23 onward
API: Input$10/1M tokens (~฿365)From June 9
API: Output$50/1M tokens (~฿1,825)From June 9
Compared2× Opus 4.8 price, <½ Mythos Preview

Free window June 9-22 for Pro/Max/Team/Enterprise subs — run heavy tests now before credits-based access starts June 23. A $20/month Pro user gets Mythos-class for no extra cost during this window.

Claude Fable 5 pricing: Input $10 Output $50 per 1M tokens, free for subscribers until June 22

Safety Fallback — Understand Before Using

Fable 5 classifies queries before answering. If a query falls into one of these 3 categories, it falls back to Opus 4.8 (not blocked):

  • Cybersecurity — exploitation, offensive cyber, agentic hacking
  • Biology / Chemistry — gene therapy design, dangerous virus engineering, bioweapon-related
  • Distillation — attempts to extract capabilities for competing models

📊 Anthropic confirms: fallback occurs in <5% of sessions — over 95% use full Fable. For normal work (coding, analytics, finance, vision), no impact.

Claude Fable 5 safety fallback: Cybersecurity / Biology / Distillation queries route to Opus 4.8

Where Thai Devs / Businesses Should Use It

Fable 5 fits these jobs best:

  • Large codebase migration / refactor — Stripe case: 2 months → 1 day. Thai companies with legacy code (Ruby/PHP/legacy Java) can test this for free right now.
  • Long-document analysis — contracts, research reports, papers. Long-context = 3× Opus 4.8 = much deeper summaries.
  • Vision OCR / diagrams — Fable 5 cleared Pokémon FireRed vision-only = reads UI screenshots, scanned documents, diagrams more accurately than prior models.
  • Finance / trading analysis — confirmed by Hebbia + IMC — Thai funds/businesses doing financial modeling can test.
  • Long agentic workflows — Anthropic states explicitly "longer tasks = bigger lead." Use cases where Opus 4.8 hits context drift are good migration targets.
  • Don't use for: short/general chat — overpriced vs Sonnet 4.6 ($3/$15 vs $10/$50).

🎯 Want to use Claude Fable 5 in a Thai business workflow? See 30 Thai business AI use cases · Claude pricing analysis · or AI Consulting.

Thai dev/business use cases: Codebase migration, Long docs, Vision OCR, Finance, Agentic

Limits + What Anthropic Hasn't Disclosed

Check these before deploying to production:

  • Context window — Anthropic hasn't published a number (Opus 4.7 has 200k, Sonnet 4.6 has 1M). Fable 5 likely similar, but wait for the official spec.
  • Thinking levels — not detailed in the announcement.
  • Deep Thai language — no Thai-specific benchmark yet. A/B test against Claude Opus 4.7 before migrating legal/medical Thai workloads.
  • 2× the price of Opus 4.8 — for tasks Opus handles well, Opus stays cheaper. Fable 5 is for "long + complex + need top quality" jobs.
  • Capacity — Anthropic flipping subs from free to credits June 23 likely means demand exceeds capacity. Test heavily this window.

Frequently Asked Questions

Claude Fable 5 คืออะไร?

Claude Fable 5 คือโมเดล AI ใหม่ของ Anthropic ที่เปิดตัว 9 มิ.ย. 2026 เป็นรุ่น Mythos-class (ระดับสูงกว่า Opus) Anthropic บอกว่า "ทรงพลังที่สุดที่เคยปล่อยให้ใช้ทั่วไป" — SOTA นโน้ม benchmark เกือบทุกตัว มี safety fallback กลับไป Opus 4.8 เมื่อเจอคำขอ cybersecurity/biology/distillation (<5% session)

Claude Fable 5 vs Mythos 5 ต่างกันยังไง?

ตัวโมเดลคือตัวเดียวกัน — Mythos 5 = full capability เปิดเฉพาะ partner cyberdefense / นักวิจัย biology ที่ผ่าน trusted access · Fable 5 = โมเดลเดียวกัน + guardrail ครบ ปล่อยให้คน public ใช้ได้ทั้งหมด ความสามารถปกติเหมือนกัน แต่ Fable จะ fallback ไป Opus 4.8 ในหัวข้ออันตราย

Claude Fable 5 ใช้ฟรีได้ไหม?

ฟรีสำหรับสมาชิก Claude Pro/Max/Team/Enterprise ระหว่าง 9-22 มิ.ย. 2026 — Pro ($20/เดือน) ก็ได้ใช้ Mythos-class ฟรีในช่วงนี้ ตั้งแต่ 23 มิ.ย. เป็นต้นไป สมาชิกต้องใช้ usage credits (จ่ายเพิ่ม) นักพัฒนาใช้ผ่าน API จ่ายตาม token เสมอ ($10 input / $50 output per 1M)

Claude Fable 5 ราคา API เท่าไหร่?

$10/1M input + $50/1M output tokens (~฿365 / ฿1,825) — 2 เท่า Claude Opus 4.8 และ น้อยกว่าครึ่งของ Mythos Preview เหมาะกับงานที่ "ยาว + ซับซ้อน + ต้องการคุณภาพสุด" สำหรับงานปกติ (chat, summarization สั้น) Sonnet 4.6 ($3/$15) คุ้มกว่าเยอะ

Claude Fable 5 เก่งกว่า Opus 4.8 ตรงไหน?

ตัวเลขทางการ: Slay the Spire (long-context + memory) 3 เท่าของ Opus 4.8 · >90% บน core analytics benchmark (เป็นโมเดลแรก) · FrontierCode คะแนนสูงสุดในกลุ่ม frontier · Hebbia Finance benchmark สูงสุด · Pokémon FireRed (vision-only) ผ่าน — รุ่นก่อนต้องใช้ helper tool · Stripe migration Ruby codebase 2 เดือน → 1 วัน Anthropic เน้น "ยิ่งงานยาว/ซับซ้อน ยิ่งทิ้งห่าง"

เมื่อไหร่ไม่ควรใช้ Fable 5?

(1) Chat สั้น/ทั่วไป — แพงเกินไป Sonnet 4.6 ทำได้พอ (2) งาน cybersecurity/biology อันตราย — fallback ไป Opus 4.8 อยู่แล้ว ไม่ได้ Fable เต็มตัว (3) Production ที่ต้องการ context window ตัวเลขชัด — Anthropic ยังไม่ระบุ รอ official ก่อน (4) Thai-specific reasoning ลึก (legal/medical) — ยังไม่มี benchmark ภาษาไทย ควร A/B test กับ Opus 4.7 ก่อน migrate

Share:
Arm - CherCode

Arm - CherCode

Full-Stack Developer & Founder

Software developer with 5+ years of experience in Web Development, AI Integration, and Automation. Specializing in Next.js, React, n8n, and LLM Integration. Founder of CherCode, building systems for Thai businesses.

Portfolio

Related Service

AI Consulting Services

Learn More