Unverified50% confidenceBenchmarkExact time
Running Gemma 4 26B on an M3 Max, MLX is 65.0 tokens/s and LLAMA.cpp is 63.9 tokens/s, essentially a tie between the two.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/13/2026
First Seen
Valid until: 1/11/2027
Sources
Related Entities
Related Claims
UnverifiedIn a fair 4-bit comparison, MLX with D-Flash 2 speculative decoding ran at 16.6 tokens/s, versus Ollama Q4's 5.7 tokens/s.76% similarUnverifiedBased on public data for high power plus MLX or Ollama, the decoding speed of different models on M4 Max is roughly a ladder as follows: Nemotron 3 Nano about 160 tokens/s, Qwen 3.5 35B at 107 to 130 tokens/s, Gemma 4 26B about 75 tokens/s, DeepSeek V4 Flash about 27 tokens/s, and dense 27B about 22 tokens/s.74% similarUnverifiedIn a llama.cpp environment with Qwen3 27B, the R9700 achieved 27.1 tokens/s (approximately 65% of bandwidth ceiling) and the B70 achieved 21.4 tokens/s (just over approximately 50%).73% similarUnverifiedJordan Ilchif tested Qwen 3.8 27B on a 32GB M4 MacBook Air, with Ollama's NVFP4 build running at 17.2 tokens/s, pure MLX at 6.6 tokens/s, and LLAMA.cpp at 4.4 tokens/s, but the formats were different.73% similarUnverifiedCommunity benchmark comparison data: M2 Ultra approximately 20 tokens/second, M3 Ultra 512 GB at 39.8 tokens/second, a single B200 at 119.7 tokens/second, 8× RTX Pro 6000 at approximately 204 tokens/second, a modified RTX 4090 with 48 GB VRAM on a custom backend at 367 tokens/second, and dual H200 at approximately 375 tokens/second.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1015151API
curl https://kongchang.com/api/v1/knowledge/claims/1015151MCP
get_claim(id=1015151)