Tasks and inputs
LongCat Flash Prover is listed for Text generation, Coding & Code Generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Oct 8, 2026
LongCat Flash Prover is a large Lean 4 reasoning checkpoint for generating and repairing formal proofs with tool feedback.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
LongCat Flash Prover is a large Lean 4 reasoning checkpoint for generating and repairing formal proofs with tool feedback.
Flash Prover is a 560B mixture-of-experts model built for Lean 4 and tool-integrated reasoning. Its useful output is a proof that the selected Lean environment accepts. A persuasive mathematical explanation alone does not meet that criterion.
Provide the theorem statement, available imports and the relevant library definitions. Compile the proposed proof, return the exact diagnostic and let the model revise it. Keep Lean and dependency versions consistent across the trial so environmental differences are not mistaken for model behaviour.
The publisher documents interleaved and preserved-thinking approaches. The ordinary template retains tool calls and responses while managing prior thinking separately. Use the supplied template and parser so compiler feedback remains part of the agent’s working context. The released configuration sets 983,040 positions, but a maximum-length trial requires substantial infrastructure.
Test several theorem types, including a statement that cannot be proved from the supplied assumptions. Record accepted proofs, tool calls, retries and total time. The MIT-licensed weight set is a large distributed deployment; active expert computation does not make the complete checkpoint small. This profile keeps ordinary API pricing unknown unless a selected host establishes a rate for its specific route.
Compare the model on the work you need to complete.
LongCat Flash Prover is listed for Text generation, Coding & Code Generation. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Documented limits: 983.04K context. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
The weights are downloadable under MIT. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Mistral AI
Leanstral 1.5 119B A6B is an Apache-licensed model for Lean 4 formal proofs and code workflows.
Mistral AI
Leanstral 2603 is an Apache-licensed model for Lean 4 formal proofs and code workflows.
inclusionAI
Ling 2.6 1T is a trillion-parameter MIT-licensed checkpoint for complex instruction following, coding and long-document workflows.
inclusionAI
Ling 2.6 Flash is a hybrid-attention instruction checkpoint designed for concise responses and repeated agent tool workflows.
inclusionAI
Ling 3.0 Flash is inclusionAI’s MIT-licensed hybrid reasoning checkpoint for long-context coding and agent tasks.
inclusionAI
Ling 3.0 Tiny is a compact MIT-licensed reasoning checkpoint with 7.9B total parameters and 1.3B active per token.