halogen 0.12.0 hits 38 tok/s decode at 1M context for Qwen on Strix Halo
peonist-ai · reddit · 2026-09-20
peonist-ai released halogen-flash-server 0.12.0, fixing context-depth performance degradation for Qwen3.8-Flash-Next on AMD's Ryzen AI Max+ 395 (128GB Strix Halo).
- Decode at 1,004,581 tokens of context improved from 27.3 to 38.3 tok/s (default speculative drafter); 42.9→45.0 tok/s at 258,794 context
- 1M prefill rose from 790 to 937 tok/s, cold-request time cut from 21.2 to 17.9 minutes; 256k prefill now 1,114 tok/s
- Follow-up turns over the prompt cache hit first token in 0.55s at 1M context
- Setup: add HALOGENROPEYARN=4 and HALOGENCTX=1048576 to the podman line; needs a 128GB box; open source on GitHub
More from Infra
- Malaysia poised to become a compute superpower with both US and Chinese chips by next year — basedjensen · 2026-11-04
- ByteDance details USDB storage system: 85.8% lower write latency for recommender sequences — _reachsumit · 2026-11-04
- Oracle cancels $7B chip leasing deal with Tencent, FT reports — SumitGup · 2026-11-04
- Inference engine shrunk to 1023 bytes, still runs SmolLM2-135M — cneuralnetwork · 2026-11-04
- Musk says space data centers are 'looking pretty appealing right now' — elonmusk · 2026-11-04
- Data centers are turning to old jet engines as the wait for power grows — IndependentCoast3974 · 2026-11-04