README.md

---
license: apache-2.0
base_model: Qwen/Qwen3.6-27B
language:
  - en
library_name: transformers
tags:
  - solidity
  - smart-contracts
  - code-generation
  - foundry
  - blockchain
  - ethereum
  - security-audit
  - rejection-fine-tuning
  - qwen
datasets:
  - ASSERT-KTH/DISL
  - braindao/solidity-base-sft-v2
  - samscrack/solidity-audit-cot
pipeline_tag: text-generation
---

# 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`](https://huggingface.co/datasets/samscrack/solidity-eval-2026) (`lite` split, 200 real Etherscan contracts):

| Agent / model | pass@1 | Wall-clock |
|---|---|---|
| **This model — Qwen 3.6 Solidity 27B** | **46.5%** (93/200) | ~27 min |
| Claude Code 2.1.128 (Claude Opus 4.7) | 39.0% (78/200, 1 timeout) | ~34 min |

`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

| # | Stage | Method | Adapter | Training data |
|---|---|---|---|---|
| 0 | Continued pretrain | LoRA r=64, ~500M Solidity tokens | folded in | `ASSERT-KTH/DISL` (514k deployed contracts, CC-BY 4.0) + ~80 curated blue-chip GitHub repos |
| 1B | Instruction SFT | LoRA r=64, 178 steps | folded in | `final.jsonl` (~315k rows: braindao/solidity-base-sft-v2 + andstor/smart_contract_code_comments + lohoz/Smart-Contract-MultiTask + slither-audited + Pyano-fun) + 4,240 unverified `foundry_tests.jsonl` rows |
| 2 | Audit / long-CoT | LoRA r=16, 2 epochs | folded in | `samscrack/solidity-audit-cot` (~6,140 Opus 4.7 long-form audit traces, all `confidence=high`, ≤30k chars to fit 8K ctx) |
| 3 | Opus distillation SFT | LoRA r=16, 2 epochs, lr=5e-5 | folded in | 4,000 of 4,919 forge-verified Opus pairs (`foundry_tests.verified.jsonl`); 919 held out from training |
| 4 | Rejection fine-tuning (RFT) | LoRA r=16, 2 epochs, lr=5e-5 | folded in (this checkpoint) | 926 model-generated contract+test pairs that passed `forge build && forge test` self-oracle, with non-triviality gate (≥3 test fns, ≥2 distinct asserts) |

**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`:

| Metric (200 prompts × N=4 candidates) | Post-Stage-3 | **Post-Stage-4** | Δ |
|---|---|---|---|
| extract success | 80.5% | **86.4%** | +5.9 pp |
| compile success | 46.8% | **50.6%** | +3.8 pp |
| test pass | 19.2% | **21.4%** | +2.2 pp |
| **prompts ≥1 pass** | 45.0% | **54.0%** | **+9.0 pp** |

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

```python
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 { ... }
```

```solidity
// 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.