Claude Fable 5.1 Is Live, at Fable 5's Price

anthropic/claude-fable-5.1 has been live on BlockRun since September 8, on Base and on Solana. It is Anthropic's most capable model, the successor to Fable 5 in the same tier, at the same price.

It takes images and tools, and thinking is always on. There is no API key and no subscription: every call is quoted in dollars before it runs and paid per call in USDC.

What it is for

Fable is the tier above Opus, for the most demanding reasoning and long-horizon agentic work: the hard bug nobody has cornered, the design decision with many moving parts, the long run where an agent has to hold a large codebase or document set in its head and still be right at the end.

The 1M context and 128K output cap matter for that kind of work. The 128K is the model's own limit, confirmed by the upstream refusing anything larger rather than copied from a page.

When it is worth the price

Fable 5.1 costs 2.5x what Opus 5.5 does, on both input and output. For most agent work that is the wrong trade: Opus 5.5 and Sonnet 5.5 cover everyday coding and multi-step agent runs for a fraction of the cost.

The case for Fable is a task where the cheaper model's failures are the expensive part — a retry loop that burns more than the difference, or an answer that has to be right the first time because a person or another system acts on it. A practical pattern is to route by difficulty: run Sonnet or Opus by default, and escalate to Fable for the steps that keep failing. On a pay-per-call gateway that is a per-request decision, not a plan change.

Thinking is always on, and thinking tokens bill as output. Expect a Fable call to produce more output tokens than the visible answer suggests, and budget max_tokens with that in mind.

The forced tool choice, and why it mattered

Fable 5.1 rejects a forced tool choice. Asking it to call a specific tool, or any tool (tool_choice: "required"), returns an error — where Fable 5 and Opus 5 both accept it. Agent frameworks send tools on almost every request, and many force the call.

On a gateway that settles payment before the upstream request, that difference is not cosmetic. Unhandled, every such request would have been paid for and then failed. The gateway now serves a forced choice as auto on this model. Fable still calls the tool in almost every case; if a call must happen, name the tool in your prompt.

Two other request rules are handled the same way:

• Sampling parameters. temperature, top_p and top_k are not accepted. The gateway drops them rather than failing the call; steer with the prompt.
• Anthropic's own model id. claude-fable-5-1, the dashed spelling Anthropic uses, resolves on both the chat endpoint and the Messages API, so code written against Anthropic's SDK does not need its model ids rewritten.

If it is unavailable

Fable 5.1 falls back to Fable 5: same maker, same tier, same $10 / $50. If the upstream for 5.1 fails, the request is served by Fable 5 at the price you were quoted. A rescue never costs more than the request you sent, and it is never a smaller model standing in for the one you chose.

Calling it

The chat endpoint is OpenAI-compatible, and the SDK handles the payment. On Base:

from blockrun_llm import LLMClient

client = LLMClient()  # wallet from BLOCKRUN_WALLET_KEY or ~/.blockrun/.session
print(client.chat("anthropic/claude-fable-5.1", "Find the race condition in this module."))

On Solana, the same call with the Solana client:

from blockrun_llm import SolanaLLMClient

client = SolanaLLMClient()  # wallet from SOLANA_WALLET_KEY; pays USDC on Solana
print(client.chat("anthropic/claude-fable-5.1", "Review this architecture for failure modes."))

If you already use Anthropic's Messages API, point it at BlockRun and keep your code: POST /v1/messages accepts both anthropic/claude-fable-5.1 and claude-fable-5-1.

Try it

The full catalog, with live prices, is at blockrun.ai/models, and the docs cover each model's request rules. Send a request, and the price comes back before the work does.