Unverified50% confidenceBenchmarkExact time
Semi Analysis tested on the first day of DeepSeek V4's release: NVIDIA's CUDA worked out of the box; the AMD platform ran V4 at only 1 to 2 words per second on the first day, and engineers spent a month fixing it to barely usable; Huawei's Ascend 950DT, with a complete deployment guide and custom operators, ran inference on the day V4 was released, the first time in history that an ecosystem other than NVIDIA's achieved day-one launch of a large new model.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/19/2026
First Seen
Valid until: 12/3/2026
Sources
Huawei Ascend's Paper Compute Beats Nvidia by 1.7x—So Why Does DeepSeek Only Trust It for Inference?
bilibili智视界AI时代11/19/2026
Related Entities
Related Claims
UnverifiedIn April 2026, DeepSeek released a preview of its first flagship V4 since R1; V4 Pro features a context window of up to one million tokens and is focused on agentic tasks.70% similarUnverifiedDeepSeek released the first visual version weights of V4, with a model scale of 305B (approximately 305 billion parameters), along with a minimal inference implementation.66% similarUnverifiedThe CS4 launched by Cerebras is claimed to be the fastest AI inference hardware currently available, ultimately capable of running models with up to 100 trillion parameters at a speed of approximately 1,000 tokens per second.65% similarUnverifiedIn the 3D monument programming test, DeepSeek V4.1 Flash produced the best overall result, with functional encapsulation and performance optimization, but took approximately one hour to complete.64% similarUnverifiedDiffuseBase released and open-sourced Clover 35BA 3B Preview, which it officially calls the first open-weight production-grade diffusion language model, using a fast diffusion mechanism, achieving the highest score among open-source models on 6 Agent benchmarks, and generating 680 tokens per second on a single card at BF16 precision, 41% faster than a same-size model accelerated by D-Flash2.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1190037API
curl https://kongchang.com/api/v1/knowledge/claims/1190037MCP
get_claim(id=1190037)