Qwen 3.6 Solidity (27B)

A 5-stage Solidity-specialist fine-tune of Qwen/Qwen3.6-27B. Trained to produce Foundry-compileable Solidity contracts and matching test suites from natural- language specs, and to reason about smart-contract security with long-CoT audit traces.

This is the final merged checkpoint — all five stages (CPT → SFT instruction → SFT audit/CoT → SFT Opus distillation → RFT) folded into a single bf16 model. Loadable directly with AutoModelForCausalLM.from_pretrained(...) — no adapters to apply.

Solidity Eval (2026) — pass@1 leaderboard

Top of the pass@1 leaderboard on samscrack/solidity-eval-2026 (lite split, 200 real Etherscan contracts):

pass@1 here is SolBench's echidna() rule: a single agentic attempt is scored 1.0 only if Diffusc compiles the candidate AND Echidna's differential-fuzz finds no behavioral divergence vs. the ground-truth body, with B3 canary + stub-residue guards. No resampling. Identical conditions across rows: 16-way concurrency, max_agent_turns=40, agent_temperature=0.6, fuzz_test_calls=50000, fuzz_seed=0xDEADBEEF, same sandbox image, same host. This model served locally via vLLM TP=2 FP8 (qwen3_xml tool parser) on 2× Blackwell GPUs through the in-process Hermes agent loop; Claude Code via Anthropic API through the CLI agent backend.

See the dataset card for the full reproduction recipe and harness-agnostic scoring instructions.

Pipeline

Stages 0/1B/2 were the original recipe (specification + Opus-CoT distillation). Stages 3/4 are the addition: directly distill the highest-quality forge-verified Opus pairs (Stage 3), then rejection-sample the model's own forge-passing outputs to anchor self-consistent generation (Stage 4).

Eval — Stage 3 → Stage 4 (RFT) comparison

200 prompts × N=4 candidates from a held-out slice (never trained on at any stage). Each model-generated (contract, test_file) pair is dropped into a fresh Foundry project and scored end-to-end with forge build && forge test:

Stage 4 RFT lifted prompt-level yield by +9 percentage points (45 → 54 %). Per-candidate compile rate jumped 10× across the full pipeline (4.5 % pre-Stage-3 → 50.6 % post-Stage-4) — the model now produces Foundry-compileable contracts with matching test suites at >50 % per individual candidate.

What this model is good at

• Producing self-consistent Foundry-compileable contract + test pairs from a NL spec. Self-oracle test pass rate is 21.4% per candidate, 54% of prompts have ≥1 of 4 passes.
• Long-CoT audit reasoning. Stage 2 was trained on ~6k Opus 4.7 audit traces with reasoning steps + structured findings (severity / category / location / impact / fix).
• Solidity-idiomatic generation. Stage 0 CPT shifts the base distribution toward modern Solidity patterns (mapping, msg.sender, pragma, custom errors, etc.).

Limitations

• Synthetic-data lineage. Stage 1B includes braindao/solidity-base-sft-v2 whose teacher model is undisclosed (likely commodity GPT, not GPT-4-class). Quality ceiling is bounded by the teacher.
• Audit-corpus legality. Stage 2 corpus (samscrack/solidity-audit-cot) is Opus-generated under Anthropic API terms over braindao seed contracts. Legal review recommended before any commercial use of the audit-finding outputs.
• Held-out eval. This model has never seen samscrack/solidity-eval-2026 (SolBench RACR-4k + differential fuzz) at any stage — that's the gold benchmark.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "samscrack/Qwen3.6-Solidity-27B",
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
)
tok = AutoTokenizer.from_pretrained("samscrack/Qwen3.6-Solidity-27B")

# Spec → contract + tests
spec = (
    "Implement a Solidity contract that holds a mapping from address to uint256 "
    "balance. Owner can mint to any address. Anyone can transfer their balance to "
    "another address. Include a Foundry test suite covering happy paths and the "
    "owner-only invariant.\n\nProduce both the Solidity contract and a Foundry "
    "test suite that exercises it."
)
msgs = [{"role": "user", "content": spec}]
inputs = tok.apply_chat_template(
    msgs, tokenize=False, add_generation_prompt=True, enable_thinking=True,
)
toks = tok(inputs, return_tensors="pt").to(model.device)
out = model.generate(
    **toks, max_new_tokens=4096, temperature=0.7, top_p=0.9, do_sample=True,
)
print(tok.decode(out[0][toks.input_ids.shape[-1]:], skip_special_tokens=True))

The generated assistant turn has the shape:

<think>...short design rationale...</think>
```solidity
// SPDX-License-Identifier: MIT
pragma solidity ^0.8.x;
contract MyContract { ... }

// test/Contract.t.sol
import "forge-std/Test.sol";
import "../src/Contract.sol";
contract MyContractTest is Test { ... }

## Format envelope

The model was trained on the canonical `<think>...</think>\n```solidity\n{contract}\n```\n\n```solidity\n// test/Contract.t.sol\n{tests}\n``` ` envelope. Most reliable
reproduction is to ask the user prompt to end with: *"Produce both the Solidity
contract and a Foundry test suite that exercises it."*

## Training infrastructure

- 2× NVIDIA RTX PRO 6000 Blackwell Workstation (96 GB each)
- Trainer: TRL 0.22 + Unsloth 2026.4.7 + PyTorch 2.8.0 + cu128
- Inference (sampling for Stage 4 RFT): vLLM 0.19.1 with FP8 dynamic quant +
  FLASH_ATTN backend + Qwen3 reasoning parser

## Citation

@misc{qwen3.6-solidity-27b, author = {Sam Crack (samscrack)}, title = {Qwen 3.6 Solidity (27B): a 5-stage CPT/SFT/RFT recipe for Foundry-compileable Solidity codegen}, year = {2026}, publisher = {HuggingFace}, url = {https://huggingface.co/samscrack/Qwen3.6-Solidity-27B} }

## License

Apache-2.0 (this checkpoint). Underlying training data is from CC-BY/MIT-tier
sources; teacher reasoning content (Stage 2 + Stage 3) was generated under
Anthropic API terms of use as of generation date (2026-05-04). Eval set
`samscrack/solidity-eval-2026` is NOT used at any training stage.