Is Claude Fable 5 available in Agent Planners?

Yes — it's selectable today on the Agents page for any of the six agents, priced at 50 credits per 1K input tokens and 250 per 1K output (≈$50/$250 per 1M at list price), on a 1,000,000-token context window.

Is Claude Opus 5 available in Agent Planners?

Yes — selectable today on the Agents page, priced at 25 credits per 1K input tokens and 125 per 1K output (≈$25/$125 per 1M at list price), the same price as Claude Opus 4.7 and 4.8, on a 1,000,000-token context window.

How does a new model become available without a product update?

Agent Planners pulls the live Vercel AI Gateway model catalog and automatically surfaces any new chat model that isn't in the hand-curated list yet, so it's selectable — and correctly priced from the real gateway rate — within hours of a provider shipping it.

Are Fable 5 and Opus 5 expensive to run?

They're premium-tier: 25-50 credits per 1K input tokens versus 1 credit for the default GPT-5.6 Luna. They're worth it for the agent where output quality drives the outcome — typically the orchestrator's planning — and overkill for high-volume, low-judgment steps.

Which agents can use Claude Fable 5 or Opus 5?

Any of the six — orchestrator, analytics, optimization, creative, safety, research. Model choice is set independently per agent on the Agents page.

Product

Claude Fable 5 and Claude Opus 5 are now in Agent Planners

TL;DR

Claude Fable 5 and Claude Opus 5 are selectable today on the Agents page, alongside every other model in Agent Planners' built-in AI Gateway. Both run on a 1M-token context window. Fable 5 prices at 50 credits per 1K input tokens and 250 per 1K output (≈$50/$250 per 1M at list price); Opus 5 at 25/125 (≈$25/$125 per 1M) — the same price as Opus 4.7 and 4.8. Neither needed us to ship a release: Agent Planners pulls the live Vercel AI Gateway catalog and auto-surfaces new chat models the moment a provider ships them.

What's new

Model	Maker	Context window	Credits / 1K (in · out)	≈ $ / 1M (in · out)
Claude Fable 5	Anthropic	1,000,000 tokens	50 · 250	$50.00 · $250.00
Claude Opus 5	Anthropic	1,000,000 tokens	25 · 125	$25.00 · $125.00

Both are Anthropic's newest generation, alongside Claude Sonnet 5 (already in the picker) and Claude Haiku 4.5. Full, always-current pricing for every model — these two included — is on the model pricing & credits ranking.

Why they showed up without a release from us

Agent Planners doesn't wait for someone to hand-edit a model list before a new model becomes usable. The Agents page pulls the live Vercel AI Gateway catalog and automatically surfaces any chat model that's live there but not yet in our curated list — labelled “(live)” until we formally add it, which we've now done for both of these.

That matters for two reasons beyond convenience. First, a genuinely new model is usable within hours of a provider shipping it, not whenever a vendor gets around to a release. Second — and this is the part that's easy to get wrong — the credit price for a brand-new model comes from the real, live gateway price as the primary source, not a hand-typed number someone forgot to add. The static fallback table only kicks in if the live catalog is briefly unreachable, and we keep it audited (more on that below).

How to use them

1Open the Agents page and pick an agent — orchestrator, analytics, optimization, creative, safety or research.
2Choose “Claude Fable 5” or “Claude Opus 5” from the model dropdown.
3Save. Usage on that agent is now metered from the live gateway price shown above — no separate API key needed, since it runs through the built-in AI Gateway.

Premium-tier pricing is real: at 25–50 credits per 1K input tokens, these are among the more expensive models in the catalog. Reserve them for the agent where quality genuinely moves the outcome — see the model-choice best-practices guide for where that pays off.

We double-check new fallback pricing, not just add the model

Adding a model to the picker is the easy part. While curating these two, we checked their real gateway price against the static fallback table (the safety net used only if the live catalog is briefly unreachable) and found Claude Fable 5 had no matching rule — it would have fallen through to a generic default that undercharges it by roughly 10–12×. Fixed before either model shipped in the curated list, with a regression test asserting the correct rate going forward.

This is the same discipline behind the GPT-5.6 Luna pricing fix — see GPT-5.6 Luna is now the default model for that story in full.

Frequently asked questions

Is Claude Fable 5 available in Agent Planners?: Yes — it's selectable today on the Agents page for any of the six agents, priced at 50 credits per 1K input tokens and 250 per 1K output (≈$50/$250 per 1M at list price), on a 1,000,000-token context window.
Is Claude Opus 5 available in Agent Planners?: Yes — selectable today on the Agents page, priced at 25 credits per 1K input tokens and 125 per 1K output (≈$25/$125 per 1M at list price), the same price as Claude Opus 4.7 and 4.8, on a 1,000,000-token context window.
How does a new model become available without a product update?: Agent Planners pulls the live Vercel AI Gateway model catalog and automatically surfaces any new chat model that isn't in the hand-curated list yet, so it's selectable — and correctly priced from the real gateway rate — within hours of a provider shipping it.
Are Fable 5 and Opus 5 expensive to run?: They're premium-tier: 25-50 credits per 1K input tokens versus 1 credit for the default GPT-5.6 Luna. They're worth it for the agent where output quality drives the outcome — typically the orchestrator's planning — and overkill for high-volume, low-judgment steps.
Which agents can use Claude Fable 5 or Opus 5?: Any of the six — orchestrator, analytics, optimization, creative, safety, research. Model choice is set independently per agent on the Agents page.