GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
CertainLogicAI
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
CertainLogicAI
Model Tree
Qwen/Qwen3.6-35B-A3B
Input Modalities
Output Modalities
Supported Functionality
| Benchmark | Base Model | +Adapter | Δ |
|---|---|---|---|
| Coder Quick (10-task) | 70.0% | 80.0% | +10.0 pp |
⚠️ McNemar p = 1.00 on 10 tasks. A single-task difference. Do not treat this as statistically significant. Proper execution-graded benchmarks (300-400 tasks) are forthcoming.
Judge Sanity (10-task, constrained format):
| Metric | Base Model | +Adapter | Δ |
|---|---|---|---|
| Accuracy | 70% | 90% | +20 pp |
| Cohen's κ | 0.40 | 0.80 | +0.40 |
| False rejects (of 5 good) | 3 | 1 | −2 |
| False accepts (of 5 bad) | 0 | 0 | — |
⚠️ n=10, McNemar p = 0.50. Direction is consistent and mechanism is clear (fewer false rejections, zero false accepts), but not yet statistically validated. A larger benchmark is the priority.
This is the single most important thing to get right. The adapter was trained on template-generated records. Without a pinned output format, it emits training-set artifacts instead of evaluating code. Use this exact schema:
json
{"judgment": "accept | reject | fix | escalate","confidence": 0.0-1.0,"issues": [{"severity": "error | warning | info","description": "What is wrong and why","line_numbers": [0]}],"strengths": ["What the code does well"],"suggested_fix": "Corrected code if applicable","explanation": "Overall assessment of code quality"}
Field semantics:
judgment: accept = passes, reject = fails, fix = minor fix needed, escalate = needs human reviewconfidence: 0.0-1.0 scale. Below 0.6 → set judgment to escalateissues: empty array if no issues foundsuggested_fix: only required when judgment is fixstrengths: at least 1 item for accept judgmentstext
You are a code evaluation assistant. Your job is to evaluate code for correctness, edge cases, and style.You MUST respond with valid JSON only, following this exact schema:{"judgment": "accept | reject | fix | escalate","confidence": 0.0-1.0,"issues": [{"severity": "error | warning | info","description": "What is wrong and why","line_numbers": [0]}],"strengths": ["What the code does well"],"suggested_fix": "Corrected code if applicable","explanation": "Overall assessment of code quality"}Rules:- judgment "accept" = code is correct. "reject" = fundamentally wrong. "fix" = minor issue. "escalate" = you are unsure.- If confidence < 0.6, set judgment to "escalate".- issues must be empty for "accept". strengths must be empty for "reject".- suggested_fix only required for "fix" judgments.- Be specific about what is wrong and why. Reference line numbers when possible.- Never output anything outside the JSON object.
bash
pip install torch transformers peft accelerate
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase_model_name = "Qwen/Qwen3.6-35B-A3B"adapter_path = "CertainLogicAI/qwen3.6-35b-a3b-the-judge-code-edition" # Note: CertainLogicAI, not CertainLogic# Load base modelmodel = AutoModelForCausalLM.from_pretrained(base_model_name,torch_dtype=torch.bfloat16,device_map="auto",)tokenizer = AutoTokenizer.from_pretrained(base_model_name)# Apply adaptermodel = PeftModel.from_pretrained(model, adapter_path)
python
import jsonsystem_prompt = """You are a code evaluation assistant. Your job is to evaluate code for correctness, edge cases, and style.You MUST respond with valid JSON only, following this exact schema:{"judgment": "accept | reject | fix | escalate","confidence": 0.0-1.0,"issues": [{"severity": "error | warning | info","description": "What is wrong and why","line_numbers": [0]}],"strengths": ["What the code does well"],"suggested_fix": "Corrected code if applicable","explanation": "Overall assessment of code quality"}Rules:- judgment "accept" = code is correct. "reject" = fundamentally wrong. "fix" = minor issue. "escalate" = you are unsure.- If confidence < 0.6, set judgment to "escalate".- issues must be empty for "accept". strengths must be empty for "reject".- suggested_fix only required for "fix" judgments.- Be specific about what is wrong and why. Reference line numbers when possible.- Never output anything outside the JSON object."""user_prompt = """Evaluate this code for correctness:```pythondef divide(a, b):return a / b
Does this code handle edge cases properly?"""
messages = [ {"role": "system", "content": system_prompt}, {"role": "user", "content": user_prompt} ]
inputs = tokenizer.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, return_tensors="pt" ).to("cuda")
outputs = model.generate(**inputs, max_new_tokens=512) result = tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True)
try: parsed = json.loads(result) print(f"Judgment: {parsed['judgment']} (confidence: {parsed['confidence']})") for issue in parsed.get('issues', []): print(f" - [{issue['severity']}] {issue['description']}") except json.JSONDecodeError: print(f"WARNING: Non-JSON output. Model may need a stricter prompt.\nRaw: {result}")
markdown
### Using as a code generator```pythonprompt = "def fibonacci(n):\n \"\"\"Return the nth Fibonacci number.\"\"\""inputs = tokenizer(prompt, return_tensors="pt").to("cuda")outputs = model.generate(**inputs, max_new_tokens=512)print(tokenizer.decode(outputs[0], skip_special_tokens=True))
The base model Qwen3.6-35B-A3B uses 256 experts with 8 routed per token plus 1 shared expert. The README on earlier versions incorrectly stated "8/32 active MoE" — the correct count is 256 experts, 8 routed + 1 shared.
The adapter was trained on q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj. Important caveat: In Qwen3.6 MoE, gate_proj, up_proj, and down_proj only target the shared expert and attention layers in 10 of 40 layers. The 256 routed expert weights (~629M adaptable parameters) were never adapted because PEFT could not hook into the fused MoE tensor structure. This means the adapter's influence is concentrated in attention projections and the shared expert, not the routed MoE experts.
The training sequence length was 2048 tokens, which is tight for code review tasks that include both the solution and the evaluation. For production use, consider increasing max_new_tokens to 4096 or more. The base model supports up to 262,144 tokens.
| Parameter | Value |
|---|---|
| Base model | Qwen/Qwen3.6-35B-A3B (34.7B params, 256 experts, 8 routed + 1 shared) |
| Method | LoRA (r=8, alpha=16, dropout=0.05) |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| Training records | 1,685 train / 81 validation |
| Data sources | 6 curated template-generated sources |
| Compute | Single H100 (80GB), ~61 minutes |
| Precision | BF16 mixed precision |
| Sequence length | 2048 (unpacked) |
Tested on 6 refusal probe cases: 0/6 false refusals. The adapter does not introduce excessive refusal behavior.
max_new_tokens for production use.Apache 2.0. The base model (Qwen3.6-35B-A3B) carries its own license terms.
Built by CertainLogic. We build deterministic AI tools for agents that execute real work.
The training dataset is proprietary and not included.
| Epochs | 1 |
| Eval loss | 0.5793 |