{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiddaxtcoqhx2qcpitnwntarmwdbpsr5v6mba732u6irsj44vbguk4",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mppen7hh3ft2"
  },
  "path": "/t/best-moa-with-models-under-40b/177373#post_1",
  "publishedAt": "2026-07-03T00:33:27.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "This combination is doable with total of around 90GB of RAM and it is superior to DeepSeek-v4-pro\nPlease see my other post.\nAny one can predict the combination:\nDrafters:\n\n  1. Gemma-4-31B-QAT-Q4 with MTP (llama.cpp)\n  2. Ornith-1.0-35B - Q4_k_M (llama.cpp)\neach with a context of 102400 and 8bit kv cache\nAggregator:\nQwen3.6-27B-4bit context 204800\n\n\n\nCall it MoA-Max:\n\nTask | MoA-MAX\n---|---\nBug Investigation | 89.2\nTool Orchestration | 89.3\nCode Architecture | 87.8\nClinical Decision | 80.4\nMulti-step Coding (DLQ) | 91.8\n**Overall** | **87.7**\n\n**Compared to single and other MoA:**\n\nRank | Model / Configuration | Bug | Tool | Arch | Clinical | DLQ | **Avg /100**\n---|---|---|---|---|---|---|---\n1 | MoA-MAX † | 89.2 | 89.3 | 87.8 | 80.4 | 91.8 | **87.7**\n2 | MoA: Gemma-12B + Ornith-35B → agg Qwen-3.6-27B † | 80.3 | 86.6 | 88.1 | 90.3 | 88.9 | **86.8**\n3 | Qwen-3.6-27B (4-bit, vLLM) | 90.8 | 86.2 | 86.2 | 80.7 | 88.9 | **86.6**\n4 | MoA: Gemma-12B + Qwen-27B → agg DeepSeek-v4-Pro | 88.1 | 88.1 | 88.1 | 81.1 | 84.2 | **85.9**\n5 | Qwen3.6-35B-A3B (Q4_K_M) | 88.1 | 83.1 | 88.1 | 77.6 | 88.9 | **85.2**\n6 | MoA: Gemma-12B + Nemotron-2 → agg Qwen-3.6-27B | 80.5 | 87.5 | 88.1 | 80.5 | 88.9 | **85.1**\n7 | MoA: Gemma-12B + Qwen-35B-A3B → agg Qwen-3.6-27B | 81.1 | 88.1 | 88.1 | 77.6 | 88.9 | **84.8**\n8 | MoA: Gemma-12B + Qwen-27B → agg Qwen3.6-35B-A3B | 88.1 | 88.1 | 86.6 | 81.1 | 77.6 | **84.3**\n9 | DeepSeek-v4-Pro (cloud) | 93.2 | 88.1 | 88.1 | 55.2 | 94.0 | **83.7**\n10 | DeepSeek-v4-Flash (cloud) | 79.5 | 88.1 | 88.1 | 73.6 | 84.9 | **82.8**\n11 | MoA: Gemma-12B + Qwen-27B → agg Ornith-35B | 80.5 | 82.4 | 84.7 | 81.1 | 84.9 | **82.7**\n12 | Ornith1.0-35B (Q4_K_M) | 73.0 | 88.1 | 86.2 | 81.1 | 84.9 | **82.7**\n13 | MoA: Gemma-12B + Qwen-27B → agg Nemotron-2 | 59.1 | 61.5 | 78.2 | 61.9 | 72.9 | **66.7**\n14 | Gemma-4-12B (4-bit, vLLM) | 69.5 | 51.5 | 56.1 | 43.9 | 49.5 | **54.1**\n15 | Nemotron-2 Cascade (Q4_K_M) | 5.8 | 5.8 | 5.8 | 5.8 | 5.8 | **5.8**\n\nSee my other post for details",
  "title": "Best MoA with models under 40B"
}