GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
FINAL-Bench
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
FINAL-Bench
Model Tree
Input Modalities
Output Modalities
Supported Functionality
| Term | Definition | Measurement |
|---|---|---|
| Model MRI | Layer-level profiling of tensor health indicators | L2 norm, Shannon entropy, std per tensor across all layers |
| LayerMRI.compare_layers | Per-tensor A vs B quality comparison yielding optimal ratio_b | score = entropy * 0.5 + std * 0.3 + clamp(norm, 100) * 0.002 per model; ratio_b = score_b / (score_a + score_b) |
| MRI-Guided Merge | Per-tensor merge ratios derived from parent diagnostics (70% MRI + 30% genome) | final_ratio = mri_ratio * 0.7 + genome_ratio * 0.3 |
| DARE-TIES | Merge algorithm: random binary mask on delta, then weighted addition | merged = A + (B - A) * random_mask(density) * ratio |
| Transplant A / B | When MRI ratio falls below 0.05 or above 0.95, one parent is used entirely | No interpolation — direct tensor copy |
| Evolutionary Search | CMA-ES population evolution over genome space (ratio, attn, ffn, embed, density_a, density_b) | Phase 1: 200 steps heuristic proxy, Phase 2: 10 steps real benchmark |
Darwin-9B-Opus is a 9B dense parameter reasoning model created using Darwin V5. Both parent models share the identical Qwen3.5-9B architecture — the Mother is a LoRA SFT on the same base, not a different architecture.
| Role | Model | Training |
|---|---|---|
| Father | Qwen/Qwen3.5-9B | Original pre-training + RLHF |
| Mother | Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled | LoRA SFT with text-only Claude 4.6 Opus reasoning chains |
Darwin V5 does not use mergekit or any external merge library. It implements DARE-TIES merge directly via PyTorch tensor operations, with MRI-guided per-layer ratios. The algorithm is inspired by the DARE-TIES method but re-implemented from scratch to support per-tensor diagnostic-guided ratios.
python
# For each tensor pair (A, B) across all safetensor shards:ta = model_a[key] # Father tensortb = model_b[key] # Mother tensor # 1. MRI diagnoses both tensorsdiag_a = LayerMRI.diagnose_tensor(ta) # {norm, entropy, std}diag_b = LayerMRI.diagnose_tensor(tb) # {norm, entropy, std} # 2. Quality score comparison determines ratio_bscore_a = diag_a["entropy"] * 0.5 + diag_a["std"] * 0.3 + min(diag_a["norm"], 100) * 0.002score_b = diag_b["entropy"] * 0.5 + diag_b["std"] * 0.3 + min(diag_b["norm"], 100) * 0.002mri_ratio = score_b / (score_a + score_b) # Higher = Mother is better # 3. Final ratio = MRI 70% + evolutionary genome 30%final_ratio = mri_ratio * 0.7 + genome_type_ratio * 0.3 # 4. DARE-TIES merge with per-tensor ratiomask = torch.rand_like(tb) < density_bdelta = (tb - ta) * maskmerged = (ta + delta * final_ratio).bfloat16()text
Phase 0: Model MRI For every tensor in both parents, measure: - L2 norm (layer energy) - Shannon entropy (weight distribution uniformity) - Standard deviation (activation spread) Compare A vs B quality scores -> per-tensor ratio prescriptionPhase 1: Evolutionary Search (200 steps, heuristic proxy) Population of 20 genomes (ratio, attn, ffn, embed, density_a, density_b) Fitness: heuristic score based on genome balance + differentiation Selection -> SLERP crossover -> Gaussian mutationPhase 2: Real Merge + Benchmark (10 steps) Top genomes from Phase 1 undergo actual tensor merge Each merge: MRI prescription (70%) + genome ratio (30%) Fitness: real benchmark score (ARC-Challenge) Best model selected and auto-uploadedPhase 3: Health Check Layer-by-layer importance comparison: child vs both parents Detect interference (child >> parents) or function loss (parents >> child)| Capability | Standard DARE-TIES | Darwin V5 |
|---|---|---|
| Implementation | mergekit library call | Direct PyTorch tensor operations |
| Ratio selection | Uniform ratio across all tensors | Per-tensor ratio from MRI diagnosis |
| Pre-merge analysis | None | Tensor-level norm/entropy/std profiling |
| Ratio determination | Human-set or grid search | MRI 70% + evolutionary genome 30% |
| Post-merge validation | Benchmark score only | Layer-by-layer child vs parents comparison |
| Architecture | Qwen3.5 Dense (Gated DeltaNet hybrid) |
| Total Parameters | 9B |
| Precision | BF16 |
| Context Length | 131,072 native |
| Languages | 201 |
| Thinking | <think> tag chain-of-thought reasoning |
| License | Apache 2.0 |
| Setup | VRAM | Status |
|---|---|---|
| BF16 Full Precision | ~20 GB | |
| NVIDIA RTX 4090 24GB | 24 GB | Comfortable |
| NVIDIA A100 40GB | 40 GB | Very comfortable |
| NVIDIA T4 16GB | 16 GB | Requires quantization |
python
from transformers import AutoTokenizer, AutoModelForCausalLMimport torch tokenizer = AutoTokenizer.from_pretrained( "FINAL-Bench/Darwin-9B-Opus", trust_remote_code=True,)model = AutoModelForCausalLM.from_pretrained( "FINAL-Bench/Darwin-9B-Opus", torch_dtype=torch.bfloat16, device_map="auto", trust_remote_code=True,) messages = [{"role": "user", "content": "Prove that sqrt(2) is irrational."}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(text, return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=4096)print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))bash
python -m sglang.launch_server \ --model-path FINAL-Bench/Darwin-9B-Opus \ --tp 1 \ --mem-fraction-static 0.90 \ --context-length 32768 \ --trust-remote-codebash
vllm serve FINAL-Bench/Darwin-9B-Opus \ --trust-remote-code \ --enforce-eager| Engine | Darwin V5 (Evolutionary Merge + Layer-Level Diagnostics) |
| Merge Method | DARE-TIES (direct PyTorch implementation, no external library) |
| MRI Integration | Per-tensor diagnosis: norm, entropy, std -> ratio prescription |
| Ratio Formula | final_ratio = mri_ratio * 0.7 + genome_ratio * 0.3 |
| Evolution | Phase 1: 200 steps proxy + Phase 2: 10 steps real benchmark |
| Best Score | 0.8508 (ARC-Challenge) |
| Infrastructure | 4 x NVIDIA H100 NVL (100GB each) |
| Developer | VIDRAFT |
| Engine | Darwin V5 |
| Base Architecture | Qwen3.5-9B |
bibtex
@misc{vidraft_darwin_9b_opus, title = {Darwin-9B-Opus: Diagnostic-Guided Evolutionary Merge}, author = {VIDRAFT}, year = {2026}, publisher = {Hugging Face}, howpublished = {\url{https://huggingface.co/FINAL-Bench/Darwin-9B-Opus}}}This model is introduced in Darwin Family.
| Transplant support |
| No |
| ratio < 0.05 -> use A entirely, ratio > 0.95 -> use B entirely |
| Failure diagnosis | "Score went down" | Per-tensor quality delta identifies problematic layers |