Unverified50% confidenceBenchmarkExact time
Flash 1017's zero-shot MMLU score is 71.2%, almost on par with GPT-4-Turbo's 71.8%, and slightly lower than DeepSeek V4-Pro's 73.5%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/17/2026
First Seen
Valid until: 10/31/2026
Sources
Related Entities
Related Claims
UnverifiedOn the DeepSWE coding leaderboard, Beam scored 44.4 points, ranking last, below DeepSeek V4.1 Flash's 74.2 points, Kimi K3's approximately 68 points, GLM's approximately 61 points, and Qwen3 Max's approximately 51 points.70% similarUnverifiedFlashMLA's prefill throughput for the sparse attention operator on the Ascend 950 reaches 410 TFLOPS (95% of the hardware's theoretical peak), and decode reaches 360 TFLOPS (83%), both figures as published by DeepSeek itself with no third-party measurements.67% similarUnverifiedFlash 1017's power consumption per million tokens is 0.42 kWh, about one-third lower than the comparison group.67% similarVerifiedDeepSeek-V4.1-Flash's KV cache is compressed to 890 bytes per token, only one-quarter that of V4 Flash.66% similarVerifiedV4 Flash's cache size is only 7% of that of V3.2.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1054136API
curl https://kongchang.com/api/v1/knowledge/claims/1054136MCP
get_claim(id=1054136)