ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
{
"schema_version": "ods.performance-evidence.v1",
"updated_at": "2026-07-18",
"entries": [
{
"id": "witcheer-4060ti-qwen35-9b-q4km-16k",
"model_id": "qwen3.5-9b",
"model_names": ["qwen3.5-9b", "Qwen3.5-9B-Q4_K_M.gguf"],
"quantization": "Q4_K_M",
"backend": "nvidia",
"gpu_name": "NVIDIA GeForce RTX 4060 Ti",
"vram_gb": 8,
"context_length": 16384,
"tokens_per_second": 44.17,
"metric": "decode_tok_s",
"runtime": "LM Studio",
"source_label": "witcheer/windows-rtx-4060ti-8gb-bench-2026-05",
"source_url": "https://huggingface.co/datasets/witcheer/windows-rtx-4060ti-8gb-bench-2026-05",
"license": "cc-by-4.0"
},
{
"id": "ods-fleet-windows-laptop-rtx5070-phi4-mini-q4km-128k",
"model_id": "phi4-mini-q4",
"model_names": ["phi-4-mini", "Phi-4-mini-instruct-Q4_K_M.gguf"],
"quantization": "Q4_K_M",
"backend": "nvidia",
"gpu_name": "NVIDIA GeForce RTX 5070 Laptop GPU",
"vram_gb": 8,
"context_length": 128000,
"tokens_per_second": 0.5,
"metric": "decode_tok_s",
"runtime": "llama-server",
"source_label": "ods-fleet/windows-laptop/model-ui-2026-07-18",
"source_url": "",
"license": "internal-fleet-validation"
},
{
"id": "witcheer-4060ti-qwen36-35b-a3b-ncmoe30-32k",
"model_id": "qwen3.6-35b-a3b",
"model_names": ["qwen3.6-35b-a3b", "Qwen3.6-35B-A3B-UD-Q4_K_M.gguf"],
"quantization": "UD-Q4_K_M",
"backend": "nvidia",
"gpu_name": "NVIDIA GeForce RTX 4060 Ti",
"vram_gb": 8,
"context_length": 32768,
"tokens_per_second": 35.36,
"metric": "decode_tok_s",
"runtime": "llama-server",
"flags": {
"n_cpu_moe": "30",
"cache_type_k": "q8_0",
"cache_type_v": "q8_0",
"flash_attn": "on"
},
"source_label": "witcheer/windows-rtx-4060ti-8gb-moe-offload-bench-2026-05",
"source_url": "https://huggingface.co/datasets/witcheer/windows-rtx-4060ti-8gb-moe-offload-bench-2026-05",
"license": "cc-by-4.0"
},
{
"id": "witcheer-4060ti-qwen36-35b-a3b-turbo3-64k-checkpoint-disabled",
"model_id": "qwen3.6-35b-a3b",
"model_names": ["qwen3.6-35b-a3b", "Qwen3.6-35B-A3B-UD-Q4_K_M.gguf"],
"quantization": "UD-Q4_K_M",
"backend": "nvidia",
"gpu_name": "NVIDIA GeForce RTX 4060 Ti",
"vram_gb": 8,
"context_length": 65536,
"tokens_per_second": 30.26,
"metric": "decode_tok_s",
"runtime": "turboquant",
"flags": {
"n_cpu_moe": "30",
"cache_type_k": "q8_0",
"cache_type_v": "turbo3",
"no_cache_prompt": "1",
"checkpoint_every_n_tokens": "-1"
},
"source_label": "witcheer/rtx-4060ti-8gb-turboquant-bench-2026-05",
"source_url": "https://huggingface.co/datasets/witcheer/rtx-4060ti-8gb-turboquant-bench-2026-05",
"license": "mit"
},
{
"id": "witcheer-4060ti-gemma4-26b-a4b-ncmoe23-32k",
"model_id": "gemma-4-26b-a4b",
"model_names": ["gemma-4-26b-a4b", "gemma-4-26B-A4B-it-Q4_K_M.gguf"],
"quantization": "Q4_K_M",
"backend": "nvidia",
"gpu_name": "NVIDIA GeForce RTX 4060 Ti",
"vram_gb": 8,
"context_length": 32768,
"tokens_per_second": 29.27,
"metric": "decode_tok_s",
"runtime": "turboquant",
"flags": {
"n_cpu_moe": "23",
"cache_type_k": "q8_0",
"cache_type_v": "q8_0",
"flash_attn": "on"
},
"source_label": "witcheer/windows-rtx-4060ti-8gb-moe-offload-bench-2026-05",
"source_url": "https://huggingface.co/datasets/witcheer/windows-rtx-4060ti-8gb-moe-offload-bench-2026-05",
"license": "cc-by-4.0"
}
]
}