Expired50% confidenceFactExact time
In the Flappy Bird HTML game generation test, the speed without MTP enabled was 31.15 tokens/second, and reached 37.3 tokens/second with MTP enabled, an improvement of approximately 20%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
DeepSeek V4 Flash MTP Speculative Decoding Real-World Test: A Guide to 20% Faster Local Inference
bilibili环球AI圈5/29/2026
Related Claims
UnverifiedAfter enabling MTP, inference speed fluctuates significantly, sometimes exceeding 40 tokens/second and sometimes dropping to 25 tokens/second.80% similarUnverifiedIn decode speed tests, speeds of approximately 45 tokens/second are achievable within 20,000 tokens, dropping sharply to approximately 20 tokens/second beyond 30,000 tokens.67% similarUnverifiedQwen 3.6 achieved a prefill speed of 400 tokens/second and a generation speed of close to 40 tokens/second during testing.67% similarUnverifiedChatGPT标准模式的输出速度通常在50-100 Token/秒左右63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/58541API
curl https://kongchang.com/api/v1/knowledge/claims/58541MCP
get_claim(id=58541)