zmzfpc/crane-next-80b

crane-next-80b

crane-next-80b is a CRANE merge: it is produced by merging two Qwen3-Next-80B-A3B checkpoints (an Instruct base and a Thinking donor) with the CRANE method — it is not trained or fine-tuned from scratch. CRANE (Constrained Reasoning Injection for Code Agents via Nullspace Editing) injects reasoning ability from the Thinking donor into the tool-disciplined Instruct / code base while preserving the base model's output format and tool-calling behavior.

Project page: [https://rpi-nsl.github.io/CRANE/](https://rpi-nsl.github.io/CRANE/) · Code: github.com/rpi-nsl/CRANE

Note: this is the CRANE weight-merging method for code agents. It is unrelated to the similarly-named "CRANE: Reasoning with constrained LLM generation" (arXiv 2502.09061), despite the shared acronym.

How it was made (CRANE)

CRANE is a training-free, parameter-editing weight merge that injects reasoning ability from a "Thinking" donor into a tool-disciplined Instruct / code base, while constraining the edit so the base model's output format and tool-calling behavior are preserved. It treats the Thinking − Instruct delta \\(\delta = \theta{\text{think}} - \theta{\text{inst}}\\) as a pool of candidate reasoning edits, and applies three composable stages per layer \\(l\\) and parameter component \\(c\\):

$$ \theta{\text{merged}}^{(l,c)} = \theta{\text{inst}}^{(l,c)} + \underbrace{\Pi{\tau,\,q(l,c)}^{\text{GSP}}}{\text{Stage 3}}\!\Big( \alpha \cdot \underbrace{S{\text{CTG}}(c,l)}{\text{Stage 2}} \cdot \underbrace{T\big(\delta^{(l,c)}\big)}_{\text{Stage 1}} \Big) $$

[image]

Three small calibration sets drive the stages — \\(\mathcal{D}R\\) (reasoning transfer), \\(\mathcal{D}A\\) (agent-behavior / tool-use preservation), and \\(\mathcal{D}_F\\) (format preservation):

• —Stage 1 — Magnitude thresholding \\(T(\delta)\\). A deterministic median-magnitude threshold keeps only the larger (top-half) delta coordinates and rescales them by 2, discarding low-confidence noise.
• —Stage 2 — Conservative Taylor Gate \\(S_{\text{CTG}}\\). From a signed, direction-aware score \\(sK(j) = -g{K,j}\,\deltaj\\) per calibration loss, CTG keeps the positive part of the per-coordinate minimum over the reasoning and agent-behavior objectives, \\(pj = [\min\{sR(j), sA(j)\}]+\\) — rewarding a coordinate only when the edit helps *both*. These aggregate into the per-component, per-layer coefficient \\(S{\text{CTG}}(c,l)\\), scaled by the single global merge strength \\(\alpha\\).
• —Stage 3 — Graduated Sigmoidal Projection (GSP). From the SVD of format-critical Instruct activations \\(Hq = Uq\Sigmaq Vq^{\top}\\), a smooth sigmoidal weight \\(\mathbf{w}q\\) (set by singular amplitude and threshold \\(\tau\\)) gives the projector \\(\Pi{\tau,q}^{\text{GSP}}(\Deltaq) = \Deltaq - \Deltaq Vq \operatorname{diag}(\mathbf{w}q) Vq^{\top}\\), attenuating high-amplitude format directions so reasoning is injected without perturbing chat-template tokens, tool-call delimiters, or JSON/schema structure.

The result is a merge that gains planning / reflection / recovery reasoning while keeping the base agent's compact, tool-call-disciplined behavior — the entire merge is a closed-form edit of the Instruct weights, with no fine-tuning.

This checkpoint's recipe

This checkpoint merges `Qwen/Qwen3-Next-80B-A3B-Instruct` (base) and `Qwen/Qwen3-Next-80B-A3B-Thinking` (donor) with:

• —Global injection strength — \\(\alpha = 0.15\\), multiplied by the per-component CTG coefficients, so the Thinking delta is added at low strength.
• —Per-layer / per-component gating — attention, expert (FFN), norm, and router components each get their own \\(S_{\text{CTG}}(c,l)\\) coefficient, varying by layer index rather than a single flat scalar.
• —Architecture-aware norm handling — Qwen3-Next's zero-centered \\((1 + w)\\) RMSNorm keeps the effective norm multiplier bounded, so norm edits are merged through the standard gated path without special clamping.
• —GSP projector — a freshly rebuilt Qwen3-Next-80B graduated-sigmoidal projector protects the format / tool-call subspace before injection.

Architecture

The merge preserves the standard Qwen3-Next-80B-A3B (hybrid MoE) topology unchanged:

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "zmzfpc/crane-next-80b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Write a Python function that returns the nth Fibonacci number."},
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))

Requires a recent transformers with Qwen3-Next support (the model was exported with transformers >= 4.57.0.dev0).

Citation / attribution

If you use this model or the CRANE method, please cite:

@misc{zhu2026crane,
  title        = {CRANE: Constrained Reasoning Injection for Code Agents via Nullspace Editing},
  author       = {Zhu, Mingzhi and Merler, Michele and Pavuluri, Raju and Patterson, Stacy},
  year         = {2026},
  eprint       = {2605.14084},
  archivePrefix= {arXiv},
  primaryClass = {cs.SE},
  url          = {https://arxiv.org/abs/2605.14084}
}

Project page: https://rpi-nsl.github.io/CRANE/ · Code: github.com/rpi-nsl/CRANE

Base models — built from two Apache-2.0 checkpoints:

• —Qwen/Qwen3-Next-80B-A3B-Instruct (base / backbone)
• —Qwen/Qwen3-Next-80B-A3B-Thinking (reasoning donor)

License: Apache-2.0 (consistent with both base models and the CRANE code).