A10

22 records across 2 models (quants: AWQ, BF16).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
A10Qwen2.5-7B AWQSGLang1544.94–––81922025-04community–source
A10Qwen2.5-7B AWQvLLM961.05–––81922025-04community–source
A10Qwen2.5-7B AWQSGLang333.00–––81922025-04community–source
A10Qwen2.5-7B AWQvLLM259.53–––81922025-04community–source
A10Qwen3 4B AWQvLLM656.60–––tg5122025-05community–source
A10Qwen3 4B AWQvLLM607.98–––tg10242025-05community–source
A10Qwen3 4B AWQvLLM469.62–––tg5122025-05community–source
A10Qwen3 4B AWQvLLM433.77–––tg5122025-05community–source
A10Qwen3 4B AWQSGLang424.58–––tg5122025-05community–source
A10Qwen3 4B AWQSGLang423.26–––tg10242025-05community–source
A10Qwen3 4B AWQSGLang413.54–––tg5122025-05community–source
A10Qwen3 4B AWQSGLang401.26–––tg5122025-05community–source
A10Qwen3 4B AWQvLLM395.12–––tg30962025-05community–source
A10Qwen3 4B AWQvLLM377.98–––tg30962025-05community–source
A10Qwen3 4B AWQSGLang369.19–––tg30962025-05community–source
A10Qwen3 4B AWQSGLang337.17–––tg30962025-05community–source
A10Qwen3 4B AWQvLLM278.73–––tg30962025-05community–source
A10Qwen3 4B AWQvLLM249.99–––tg30962025-05community–source
A10Qwen3 4B AWQSGLang239.82–––tg30962025-05community–source
A10Qwen3 4B AWQSGLang238.21–––tg30962025-05community–source
A10Qwen3 4B BF16vLLM138.3––––2026-10-01community–source
A10Qwen3 4B BF16vLLM132.6––––2026-10-01community–source

All rows cite their source. Provenance: community 40018 sourced 1085

← all hardware