---
title: "Tiel Repacks Ornith-1.5 to Match Claude Opus on Local Coding"
subtitle: "A community re-quantization of Ornith-1.5-35B-A3B revives its multi-token prediction head, matching Opus 4.6 on SWE-bench-Live at 23 GB."
url: "https://alphasignal.ai/news/tiel-repacks-ornith-1-5-to-match-claude-opus-on-local-coding"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-08-24T18:51:06.000Z"
updated: "2026-09-21T03:39:47.738Z"
type: "Model"
topics:
  - "Development"
  - "Gpus"
subtopics:
  - "Code Agents"
  - "Quantization"
upvotes: 3944
source_url: "https://huggingface.co/peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF-MTP"
comments: 0
truncated: true
---

# Tiel Repacks Ornith-1.5 to Match Claude Opus on Local Coding

> A community re-quantization of Ornith-1.5-35B-A3B revives its multi-token prediction head, matching Opus 4.6 on SWE-bench-Live at 23 GB.

# Tiel repacks Ornith for local coding and faster decoding

[Tiel model card](https://huggingface.co/peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF-MTP) packages [Ornith-1.5](https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B) as llama.cpp-compatible GGUF files with a repaired multi-token-prediction head, a new importance matrix, and a terser chat template called Sharp. The community release requires no additional fine-tuning and is available under the MIT license inherited from Ornith.

The model card reports 12 fixes across a 25-problem SWE-bench-Live sample, matching Opus 4.6 at its medium setting. Tiel ran at 4-bit precision from a 22.7 GB file, placing it within reach of local systems with 24 or 32 GB of combined GPU and CPU memory capacity.

__Reported SWE-bench-Live results on 25 problems__
| Model                                                                     | Problems fixed | Fix rate |
| ------------------------------------------------------------------------- | -------------- | -------- |
| Tiel                                                                      | 12             | 48%      |
| Opus 4.6, medium                                                          | 12             | 48%      |
| [Nail](https://huggingface.co/peculiar-ragdoll/Nail-Qwen3.6-35B-A3B-GGUF) | 9              | 36%      |
| Ornith-1.5                                                                | 8              | 32%      |
| Sonnet 5, medium                                                          | 8              | 32%      |

A 25-task sample can expose meaningful regressions or gains, but it cannot establish a broad ranking among models. The reported comparison also lacks an ablation that isolates the effects of the template, quantization, and other evaluation settings.

## How the draft head was repaired

Multi-token prediction, or MTP, adds a small head that drafts future tokens while the main model verifies them. Accepted drafts let llama.cpp process several tokens during a forward pass, reducing the number of sequential decoding steps.

Ornith’s initial quantization ladder contained an untrained MTP block. Every matrix had a standard deviation of 0.020 and kurtosis of 3.00, a distribution consistent with fresh random initialization. Such weights produce unusable drafts, so the head was removed from that release.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Tiel-Coder-35B-A3B-GGUF-MTP model card](https://huggingface.co/peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF-MTP)
- [Ornith-1.5-35B-A3B base model](https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B)
- [Nail-Qwen3.6-35B-A3B-GGUF](https://huggingface.co/peculiar-ragdoll/Nail-Qwen3.6-35B-A3B-GGUF)