Inco Splash hits 144 tok/s on Qwen3.8-27B M5 Max, 3x faster than Ollama
ResearchCrafty1804 · reddit · 2026-09-19
Open-source inference engine Inco Splash, built for Apple Silicon, runs Qwen3.8-27B at 144 tok/s on an M5 Max — up to 3x Ollama and 2x oMLX, 4x with agent fan-out. One-command setup, works with Claude Code, OpenCode, Codex, and LM Studio.
More from coding & agent
- Mnemos MCP nears launch after shared memory layer fix — RileyRalmuto · 2026-11-04
- Moving mobile dev to remote machines makes it multiplayer across product, design, engineering — rudrank · 2026-11-04
- Kuaishou's EvoRec: self-evolving agentic recommender lifts ad revenue 4.6% — _reachsumit · 2026-11-04
- TRIO: three specialized agents with aligned objectives refine retrieval-augmented generation — _reachsumit · 2026-11-04
- How Multica uses Multica to build itself: full human-agent dev workflow — jiayuan_jy · 2026-11-04
- Claude Code fixes game copyright refusal rule in 2.1.317: naming a game now works — apples_jimmy · 2026-11-04