Sync 2026-10-23: 169 models

README.md

@@ -36,7 +36,7 @@ table at **[modelfit.io/data](https://modelfit.io/data/)**; the canonical
 machine-readable source is
 **[modelfit.io/api/dataset](https://modelfit.io/api/dataset/)**.
 
-- **161 models** across 27 families (121 with a registry-verified local build, 23 open-weight models without a consumer-tier Ollama build, 17 closed cloud APIs tracked for comparison)
+- **169 models** across 32 families (126 with a registry-verified local build, 26 open-weight models without a consumer-tier Ollama build, 17 closed cloud APIs tracked for comparison)
 - **License:** CC BY 4.0: reuse freely, including commercially, with attribution
 - **Updated:** see the `updated` field in `models.json` (regenerated from the live endpoint)
 

models.csv

@@ -160,3 +160,11 @@ Ministral 3 3B,Mistral,3.4,Q4_K_M,6,3,,true,true,false,ollama|llama.cpp|lm-studi
 Ministral 3 8B,Mistral,8.4,Q4_K_M,12,6,,true,true,false,ollama|llama.cpp|lm-studio,"Chat, Coding, Vision",ollama run ministral-3:8b
 Ministral 3 14B,Mistral,13.5,Q4_K_M,16,9.1,,true,true,false,ollama|llama.cpp|lm-studio,"Chat, Reasoning, Vision",ollama run ministral-3:14b
 Inkling Small,Inkling,276,API,0,0,,false,true,true,,"Agentic coding, Multimodal, Reasoning",
+Flint 40B-A2.5B,Flint,39.8,NVFP4,36,22.9,28,true,true,false,ollama|llama.cpp|lm-studio,"Agentic coding, Long context",ollama run flint:40b
+Mellum2.1 12B-A2.5B Thinking,Mellum,12.1,Q4_K_M,16,8.1,14,true,true,false,ollama|llama.cpp|lm-studio,"Agentic coding, Small machines",
+Kolibri 1,Kolibri,78.1,Q4_K_M,96,49.8,20,true,true,false,ollama|llama.cpp|lm-studio,"Reasoning, German, Long context",
+Kolibri 1 (IQ4_XS),Kolibri,78.1,IQ4_XS,64,43.6,20,true,true,false,ollama|llama.cpp|lm-studio,"Reasoning, German, Long context",
+AREX-2 27B,AREX,27.4,Q4_K_M,32,17.2,64,true,true,false,ollama|llama.cpp|lm-studio,"Agents, Deep research, Coding",
+Muse Spark Open,Muse,816,API,0,0,,false,true,false,,"Agentic coding, Multimodal, Long context",
+MiniMax M3.1 Flash,MiniMax,428,API,0,0,,false,true,false,,"Coding, Agentic, Computer use",
+Beam 1.0,Beam,501,API,0,0,,false,true,false,,"Agentic coding, Reasoning, Long context",

models.json

@@ -11,12 +11,12 @@
     "note": "A model fits a device when its estimatedLoadGb is within the memory budget: ~70% of unified memory up to 32GB, scaling linearly to ~85% at 128GB and above (high-RAM Macs can wire more memory to the GPU). Q4_K_M needs roughly 0.6 GB per billion parameters.",
     "fieldSemantics": "runsLocally = a registry-verified Ollama build fits at least one consumer RAM tier tracked here (up to 256GB). openWeights = the weights are publicly downloadable. Some open-weight giants (e.g. NVIDIA Nemotron 3 Ultra, 550B, ~190GB at 2-bit) are openWeights: true but runsLocally: false because no consumer tier fits them."
   },
-  "updated": "2026-10-21",
+  "updated": "2026-10-24",
   "counts": {
-    "models": 161,
-    "local": 121,
-    "cloud": 40,
-    "families": 27
+    "models": 169,
+    "local": 126,
+    "cloud": 43,
+    "families": 32
   },
   "models": [
     {
@@ -2917,6 +2917,146 @@
       "runtimes": [],
       "bestFor": "Agentic coding, Multimodal, Reasoning",
       "ollamaCommand": null
+    },
+    {
+      "model": "Flint 40B-A2.5B",
+      "family": "Flint",
+      "params": 39.8,
+      "quantization": "NVFP4",
+      "minRamGb": 36,
+      "estimatedLoadGb": 22.9,
+      "kvKbPerToken": 28,
+      "runsLocally": true,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [
+        "ollama",
+        "llama.cpp",
+        "lm-studio"
+      ],
+      "bestFor": "Agentic coding, Long context",
+      "ollamaCommand": "ollama run flint:40b"
+    },
+    {
+      "model": "Mellum2.1 12B-A2.5B Thinking",
+      "family": "Mellum",
+      "params": 12.1,
+      "quantization": "Q4_K_M",
+      "minRamGb": 16,
+      "estimatedLoadGb": 8.1,
+      "kvKbPerToken": 14,
+      "runsLocally": true,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [
+        "ollama",
+        "llama.cpp",
+        "lm-studio"
+      ],
+      "bestFor": "Agentic coding, Small machines",
+      "ollamaCommand": null
+    },
+    {
+      "model": "Kolibri 1",
+      "family": "Kolibri",
+      "params": 78.1,
+      "quantization": "Q4_K_M",
+      "minRamGb": 96,
+      "estimatedLoadGb": 49.8,
+      "kvKbPerToken": 20,
+      "runsLocally": true,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [
+        "ollama",
+        "llama.cpp",
+        "lm-studio"
+      ],
+      "bestFor": "Reasoning, German, Long context",
+      "ollamaCommand": null
+    },
+    {
+      "model": "Kolibri 1 (IQ4_XS)",
+      "family": "Kolibri",
+      "params": 78.1,
+      "quantization": "IQ4_XS",
+      "minRamGb": 64,
+      "estimatedLoadGb": 43.6,
+      "kvKbPerToken": 20,
+      "runsLocally": true,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [
+        "ollama",
+        "llama.cpp",
+        "lm-studio"
+      ],
+      "bestFor": "Reasoning, German, Long context",
+      "ollamaCommand": null
+    },
+    {
+      "model": "AREX-2 27B",
+      "family": "AREX",
+      "params": 27.4,
+      "quantization": "Q4_K_M",
+      "minRamGb": 32,
+      "estimatedLoadGb": 17.2,
+      "kvKbPerToken": 64,
+      "runsLocally": true,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [
+        "ollama",
+        "llama.cpp",
+        "lm-studio"
+      ],
+      "bestFor": "Agents, Deep research, Coding",
+      "ollamaCommand": null
+    },
+    {
+      "model": "Muse Spark Open",
+      "family": "Muse",
+      "params": 816,
+      "quantization": "API",
+      "minRamGb": 0,
+      "estimatedLoadGb": 0,
+      "kvKbPerToken": null,
+      "runsLocally": false,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [],
+      "bestFor": "Agentic coding, Multimodal, Long context",
+      "ollamaCommand": null
+    },
+    {
+      "model": "MiniMax M3.1 Flash",
+      "family": "MiniMax",
+      "params": 428,
+      "quantization": "API",
+      "minRamGb": 0,
+      "estimatedLoadGb": 0,
+      "kvKbPerToken": null,
+      "runsLocally": false,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [],
+      "bestFor": "Coding, Agentic, Computer use",
+      "ollamaCommand": null
+    },
+    {
+      "model": "Beam 1.0",
+      "family": "Beam",
+      "params": 501,
+      "quantization": "API",
+      "minRamGb": 0,
+      "estimatedLoadGb": 0,
+      "kvKbPerToken": null,
+      "runsLocally": false,
+      "openWeights": true,
+      "ggufDiy": false,
+      "runtimes": [],
+      "bestFor": "Agentic coding, Reasoning, Long context",
+      "ollamaCommand": null
     }
   ]
 }