Research firm Artificial Analysis found V4-Flash costs 3 cents per test run, versus $3.15 for Anthropic's Claude Fable 5
NurPhoto / Getty Images
DeepSeek officially released its V4-Flash model on Friday, and research firm Artificial Analysis found it is the least expensive well-known AI model to run globally — costing roughly 3 cents per benchmark test compared with 86 cents for Moonshot AI's Kimi K3, $1.86 for OpenAI's GPT-5.6 Sol, and $3.15 for Anthropic's Claude Fable 5, according to Reuters.
For V4-Flash, DeepSeek's published rate is $0.14 for every million input tokens and $0.28 for every million output tokens. Artificial Analysis favors cost-per-test over sticker price because the two figures can diverge sharply — a model that looks cheap per token may still rack up a large bill if it needs many more steps to reach a correct answer.
In terms of performance, Artificial Analysis's Intelligence Index placed V4-Flash at 50 out of 100, a composite drawn from nine separate benchmarks that test coding ability, reasoning, and practical workplace scenarios. V4-Flash ties Google $GOOGL's Gemini 3.6 Flash and sits a single point below both Meta $META's Muse Spark 1.1 and Z.AI's GLM-5.2. Kimi K3 from Moonshot reached 57, and the gap widens further up the leaderboard, where Anthropic's Claude Opus 5 and Fable 5 alongside OpenAI's GPT-5.6 all outpace V4-Flash by at least nine points.
A higher-end follow-up model, V4-Pro, is in the pipeline, though DeepSeek has yet to say when it will officially ship.
The Chinese startup's R1 model became a global sensation in early 2025, triggering a selloff in technology stocks and prompting questions about large U.S. AI infrastructure investments. Since then, rivals including Moonshot AI, MiniMax, ByteDance, and Alibaba have moved aggressively into the space, eroding DeepSeek's early dominance. DeepSeek offered a 75% promotional discount on V4-Pro earlier this year, and both V4 models were designed to run on Huawei's Ascend AI processors in addition to Nvidia $NVDA hardware.
The Next Web reports that DeepSeek has completed its first external fundraise, bringing in upward of $7 billion — capital that should allow the company to keep undercutting rivals on price as it fights for broader adoption. OpenAI has cut GPT-5.6 pricing in response to competitive pressure, reflecting a broader market trend toward lower costs at a rapid pace.
The Next Web notes that V4-Flash targets use cases where scale is the dominant concern — think customer-facing chatbots, developer tooling, and automated back-office workflows that generate millions of requests and turn even small per-unit savings into significant cost differences.
Separately on Monday, Alibaba unveiled its Qwen3.8-Max, described as its largest and most capable AI model to date.
Join 500,000+ readers who start their day with Quartz.
By subscribing, you agree to our Terms of Service and Privacy Policy.
