DeepSeek officially released its V4-Flash model Friday, pricing it at $0.14 per million input tokens and $0.28 per million output tokens, a rate research firm Artificial Analysis says makes it by far the least expensive well-known AI model to run globally, more than 100 times cheaper than Anthropic’s Claude Fable 5 on a benchmark-cost basis. A price gap this wide between two publicly available frontier-adjacent models is what NewsTrackerToday traces round as the more consequential number than either model’s specific capabilities.
The comparison Artificial Analysis actually ran matters more than the headline per-token pricing: the firm estimated V4-Flash’s average cost at 3 cents per test on its benchmark suite, against 86 cents for Moonshot AI’s Kimi K3, $1.86 for OpenAI’s GPT-5.6 Sol, and $3.15 for Claude Fable 5. That per-test figure accounts for how many tokens a model actually needs to process and generate to complete a task, a more realistic value measure than sticker price alone, since a cheap-per-token model that needs many more steps to reach an answer can still end up costing more in practice.
Liam Anderson reads the capability tradeoff sitting underneath that price gap: “V4-Flash scored 50 out of 100 on Artificial Analysis’s Intelligence Index, tied with Google’s Gemini 3.6 Flash and just behind Meta’s Muse Spark 1.1 and Z.AI’s GLM-5.2. Kimi K3 scored 57, while Anthropic’s Claude Opus 5, Claude Fable 5, and OpenAI’s GPT-5.6 all scored nine or more points higher still. That’s a real, measurable capability gap, not just a pricing story, DeepSeek isn’t claiming V4-Flash matches the frontier, it’s explicitly trading capability for cost at a specific, deliberate point on that curve.” That explicit tradeoff, more than the price alone, is what NewsTrackerToday weighs to as the more accurate way to read this release.
DeepSeek’s own history makes this release read as a deliberate return to form: the company’s R1 model became a global sensation in early 2025, triggering a selloff in technology stocks and raising real questions about how much U.S. companies needed to spend on AI infrastructure at all, before DeepSeek got “quickly besieged,” in the report’s words, by domestic rivals including Moonshot, MiniMax, Z.AI, and tech giants ByteDance and Alibaba.
Daniel Wu, who covers geopolitics and energy, reads the competitive landscape this release lands inside: “Sources say DeepSeek is preparing for a potential IPO even as it releases V4-Flash specifically to reclaim the low-cost positioning that made it famous in the first place. The timing also isn’t isolated, Alibaba unveiled its largest model yet, Qwen3.8-Max, the same Monday, meaning China’s AI ecosystem is producing genuine competitive intensity domestically, not just against U.S. labs. That internal competition is arguably doing more to drive down prices globally than any single company’s release.” That domestic competitive pressure, more than DeepSeek’s specific pricing decision, is what NewsTrackerToday checks on as the more durable force shaping where AI pricing heads next.
DeepSeek is also preparing a more capable version of the model, called V4-Pro, though the company hasn’t given a release date, suggesting V4-Flash is deliberately positioned as the low-cost, lower-capability entry point in a broader two-tier product strategy rather than DeepSeek’s full technical answer to frontier labs.
None of this confirms whether enterprises actually shift meaningful workloads toward a 100-times-cheaper option once the roughly nine-to-fifteen-point capability gap against top-tier models factors into real deployment decisions, since plenty of use cases genuinely require the higher capability tier regardless of cost. Whether V4-Flash’s pricing forces further cuts across the rest of the industry the way R1 did in early 2025, or whether this round of Chinese price competition settles into a stable, capability-segmented market instead, is what News Tracker Today reads to as the real question this release leaves open.
