[Update Benchmark] Basalt's new PRs (Strata fork) new record IQ3_S Qwen3.8 Flash Next on single RTX 5080 and 51 GB RAM with 5500 tok/s prompt processing and up to 275 tok/s peak generation (during opencode coding live session)
Basalt's new PRs (Strata fork) achieved a new record for IQ3_S Qwen3.8 Flash Next on a single RTX 5080 with 51 GB RAM. This setup demonstrated 5500 tok/s prompt processing and peak generation speeds of up to 275 tok/s during an opencode coding live session. Specifically, code and tool calls reached 275 tok/s, while reasoning traces hit 190 tok/s. The hardware used included a 5080 16GB PCIe gen5x16, 64 DDR5 6400, and a Core Ultra 9 285K.
Time & source
Times shown in UTC
Display time zone: UTC
Local time zone unavailable; showing UTC.
PublishedOffset at this time: UTC+0Oct 29, 2026, 03:57 UTC
IngestedOffset at this time: UTC+0Oct 29, 2026, 15:01 UTC
- Published
- Oct 29, 2026, 03:57
- Ingested
- Oct 29, 2026, 15:01
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Tier is a per-source editorial setting, not a per-item score.
Code and tool calls can peak up to 275 tok/s - Reasoning traces up to 190 tok/s
Some PRs for the optimizations under my name are still pending (xyehya) but upstream already have solid gains & memory optimizations. and yes the IQ3_S is super usable.