witcheer/gemma-3-27b-eagle3-drafter
0.7B • Updated
🏗️ Building on HF
witcheer
witcheer
AI & ML interests
Local AI Maxxing.
Recent Activity
updated a model about 8 hours ago
witcheer/gemma-3-27b-eagle3-drafter published a model about 8 hours ago
witcheer/gemma-3-27b-eagle3-drafter updated a dataset 1 day ago
witcheer/rtx-5090-benchmarksOrganizations
updated a
model about 8 hours ago
published a
model about 8 hours ago
updated a
dataset 1 day ago
updated a
dataset about 1 month ago
published a
dataset about 1 month ago
updated a
dataset about 1 month ago
updated a
bucket 2 months ago
published a
bucket 2 months ago
updated a
dataset 2 months ago
published a
dataset 2 months ago
upvoted 3
changelogs 2 months ago
Hugging Face Changelog
Copy Repo Contents to Buckets Instantly
• 97
Hugging Face Changelog
Filter Models page by Base Models only
• 188
Hugging Face Changelog
Publish models from CI without HF_TOKEN
• 135
published a
dataset 2 months ago
updated a
model 2 months ago
published a
model 2 months ago
updated a
model 2 months ago
published a
model 2 months ago
updated a
dataset 2 months ago
posted an update 2 months ago
Post
260
new dataset: which local LLM best *drives an agent*? benchmarked 4 models for pairing with Hermes Agent (@NousResearch ) - a CodeAct agent that writes python to call its tools. RTX 5090, llama.cpp. two phases, hybrid:
>>> phase A (synthetic): scored 4 axes — code-as-action, long-context, instruction-following under Hermes' real ~3.5K-token prompt, multi-step loops. top was a near-tie (within noise): an 18B frankenmerge (Qwopus) edged Qwen3.6-27B, and Hermes' own 36B came LAST.
>>> phase B (real harness): installed Hermes, ran the top 3 through 14 multi-step tasks x3 repeats. the tie broke — and an efficiency gap appeared:
Qwen is perfect AND 2.4-3.7x more token-efficient — something a synthetic test can't see (only the real agent loop can). verdict: Qwen3.6-27B for local Hermes.
dataset: witcheer/hermes-pairing-bench
collection: witcheer/rtx-5090-benchmark-rig-6a17e365b534abb474250e11
>>> phase A (synthetic): scored 4 axes — code-as-action, long-context, instruction-following under Hermes' real ~3.5K-token prompt, multi-step loops. top was a near-tie (within noise): an 18B frankenmerge (Qwopus) edged Qwen3.6-27B, and Hermes' own 36B came LAST.
>>> phase B (real harness): installed Hermes, ran the top 3 through 14 multi-step tasks x3 repeats. the tie broke — and an efficiency gap appeared:
Qwen3.6-27B 100% | 3.0 turns | 364 tok
Qwopus-18B 85.7% | 3.6 turns | 870 tok
Nemotron-30B 85.7% | 4.4 turns | 1334 tokQwen is perfect AND 2.4-3.7x more token-efficient — something a synthetic test can't see (only the real agent loop can). verdict: Qwen3.6-27B for local Hermes.
dataset: witcheer/hermes-pairing-bench
collection: witcheer/rtx-5090-benchmark-rig-6a17e365b534abb474250e11