ResearchArticle en anglais
DeepSeek Vision: Does V4 Support Image Input?
DeepSeek V4 is text-only in the public API today. The Visual Primitives report shows DeepSeek's vision research, but not a generally available vision endpoint.
Lire l’articleActualités DeepSeek
Suivez les actualités des modèles DeepSeek, de l’API, des tarifs et des outils pour développeurs, ainsi que les nouvelles de l’IA. DeepSeek V4 Hub est une publication indépendante.
Dernières actualités
135CommunityArticle en anglais
Community data shows DeepSeek V4 cache hit rate above 90%: combined with the cache-hit price cut, effective input cost drops by roughly 80%
Community testing confirms that DeepSeek V4's context caching delivers 90%+ hit rates in real workloads. With nine out of ten input tokens now billed at one-tenth the original cache price, the effective blended input cost lands around 80% below the pre-cut baseline — far more than the headline price cut alone would suggest.
Lire l’articleCommunityArticle en anglais
DeepSeek V4 Flash Runs on 128GB Macs as a Lab Route
Fresh community work lowers the first credible Mac target from only ultra-high-memory experiments to a narrower 128GB path: a new ~90 GB antirez IQ2XXS quant, a dedicated fork, and a reproducible bootstrap script. This remains a community workflow, not official Mac support.
Lire l’articleOfficialArticle en anglais
DeepSeek V4 Pro Price Cut Goes Live at About $0.90
DeepSeek has turned pricing into the top story: V4 Pro now sits around $0.90 equivalent in headline pricing, while cache-hit input drops to RMB 0.025 per 1M tokens. For repeated coding and agent workloads, the effective bill can fall much lower than the headline rate.
Lire l’articleResearchArticle en anglais
DeepSeek V4 benchmark watch: Pro and Flash need separate score tracking
Independent model trackers are beginning to matter more than launch claims because DeepSeek V4 now has two different routes: Pro for harder reasoning and Flash for high-volume cost-performance.
Lire l’articleCommunityArticle en anglais
DeepSeek V4 Flash Mac and GGUF Route Becomes Reproducible
Community work around DeepSeek V4 Flash has moved from screenshots into reproducible local-deployment notes: GGUF packaging, compatible llama.cpp branches, and Mac memory limits are now the key watch points.
Lire l’articleMarketArticle en anglais
DeepSeek V4 Cost Pressure Shifts to Deployment Economics
The post-launch DeepSeek V4 story is moving toward deployment economics: long context, Flash pricing, open weights, and hardware strategy are now the angles that matter for developers and buyers.
Lire l’articleOfficialArticle en anglais
DeepSeek V4 preview goes live: 1M context, Pro/Flash variants, API available today
DeepSeek's April 24 official release makes V4 the new flagship family with 1M context, named Pro and Flash variants, and pricing that has since been cut further. V4-Pro is now listed at $0.435/$0.87 per 1M input/output tokens, with those rates becoming the official quarter-price baseline after the 75% discount window ends on May 31, 2026.
Lire l’articleLeakArticle en anglais
Aggregated DeepSeek V4 specs: ~1T MoE, Engram memory, 1M+ context
Third-party explainer sites deepseekv4.dev and deepseek-v4.ai aggregate the leaked spec sheet: ~1T-parameter MoE, Engram memory core, mHC hyper-connections reasoning, and a 1M+ token context window.
Lire l’articleLeakArticle en anglais
DeepSeek V4 launch roundup: 2M context, dynamic sparse, Ascend-trained
Cross-platform chatter on Twitter, Reddit and Hacker News converges on three claims about DeepSeek V4: a 2,000,000-token context window, a dynamic sparse MoE architecture, and end-to-end pre-training on Huawei Ascend.
Lire l’articleOfficialArticle en anglais
DeepSeek-V3 RMSNorm Kernel Optimization Commit
A new commit titled "act_quant_kernel" merged into the DeepSeek-V3 repository, indicating ongoing low-level kernel optimizations for RMSNorm operations.
Lire l’articleOfficialArticle en anglais
DeepSeek Fixes RMSNorm and Activation Quantization Kernel
DeepSeek's official GitHub repository has committed fixes for RMSNorm and activation quantization kernels, indicating ongoing low-level optimization work.
Lire l’article