StepFun Prover Preview 7B
StepFun Prover Preview 7B is a theorem-proving model derived from DeepSeek-R1-Distill-Qwen-7B. It refines a proof sketch through interaction with Lean 4.
Research release
Use Lean feedback in the loop
The documented prompt wraps partial solutions in sketch tags and returns Lean REPL feedback in a separate tagged block. The inference system must run that feedback loop; ordinary chat generation does not perform verification.StepFun StepFun-Prover-Preview-7B model card
Keep the prover and verifier separate
Practical guidance. Use the model to propose proof steps and Lean to check them. Record the exact library version and imports with each result. A proof that depends on a different environment may fail when moved to your project.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | StepFun-Prover-Preview-7B |
| Base model | DeepSeek-R1-Distill-Qwen-7B.StepFun StepFun-Prover-Preview-7B model card |
| License | Apache 2.0 is declared in the model card.StepFun StepFun-Prover-Preview-7B model card |
| Access | Published research-preview weights.StepFun StepFun-Prover-Preview-7B model card |
Related models
- StepFun Formalizer 7B
StepFun · StepFun Formalizer
StepFun Formalizer 7B converts natural-language mathematical problems into formal Lean 4 statements.
Launch date not confirmedResearch releaseDownloadable - StepFun Prover Preview 32B
StepFun · StepFun Prover
A DeepSeek-R1-Distill-Qwen-32B-based model for iterative Lean 4 proof search.
Launch date not confirmedResearch releaseDownloadable