Unverified50% confidenceBenchmarkExact time
In the 16-concurrency, 256-token-output test, Gemma 4 E2B achieves an output speed of 1887 tokens/second, and the models faster than it include Qwen2.5-0.5B (4176, 2.21x), Qwen3-0.6B (3210, 1.70x), Qwen2.5-1.5B (2780, 1.47x), and Qwen3-1.7B (2049, 1.09x).
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
单块TPU v5e实测：哪些小模型能与Gemma 4 E2B一战？
Related Entities
Related Claims
UnverifiedIn the scenario where 16 requests each carry a 3,072-token long prompt, Gemma 4 E2B maintains 1049 tokens/second, while Granite 3.3 2B drops to 732 (0.70x), SmolLM3-3B drops to 725 (0.69x), Phi-4-mini drops to 404 (0.39x), and Qwen3-4B reaches only 373 (0.36x).76% similarUnverifiedIn Strata's speed table, under a 128k context, Q2_0 (2-bit) achieves 65.1 tokens/s and IQ2_XS achieves 52 tokens/s; under a 4k short context, IQ3_XXS achieves 65.6 tokens/s.76% similarUnverifiedQwen3 27B achieved a 100% pass rate in the 256K long-context memory retrieval test (needle in a haystack), with all 15 runs being accurate.73% similarUnverifiedQwen 3.8 has a base speed of 22 tokens/s on the M5 Max, which can be boosted to 34 or even 56 tokens/s with MTP enabled.73% similarUnverified在本地推理速度对比中，Agent A1（95 tokens/秒）快于Gemma 4 Coder和Qwen系列73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1014581API
curl https://kongchang.com/api/v1/knowledge/claims/1014581MCP
get_claim(id=1014581)