Best local LLM for English

Measured English scores of quantized local models, per GPU. Highest average so far: gemma4-26b-a4b-qat on AMD Radeon RX 9070 XT.

Model	GPU	Decode tok/s	Knowledge (translated) %	Local exams %	Math %	Translation chrF++	Source
`gemma4-26b-a4b-qat`	AMD Radeon RX 9070 XT	90	82.2	-	96.8	-	maintainers
`qwen3.8-27b-q3`	AMD Radeon RX 9070 XT	50	81.5	-	94.4	-	maintainers
`qwen3.5-9b-q4`	AMD Radeon RX 9070 XT	42	80.5	-	93.6	-	maintainers
`qwen3.5-4b-q4`	AMD Radeon RX 9070 XT	78	75.5	-	91.6	-	maintainers
`gemma4-e4b-q4`	AMD Radeon RX 9070 XT	105	71.5	-	92.4	-	maintainers
`qwen3.8-27b-iq2`	AMD Radeon RX 9070 XT	40	74.2	-	-	-	maintainers

Zero-shot, thinking off; margins about ±4-5 points. How each column is measured: leaderboard. Missing your GPU? Run localllm eval --langs en,en and share it with localllm report.