DeepSeek V4 Flash is 90% off through Novita on AI Gateway

The announcement introduces a temporary pricing promotion for the DeepSeek v4 Flash model when accessed via Vercel’s AI Gateway. The promotion is gated behind the Vercel Pro subscription tier, meaning only those customers can benefit. To activate the discount, developers need to specify the model identifier and set Novita as the primary provider in the gateway order array. This change is purely a configuration tweak; the underlying model and API remain unchanged.

Financially, the discount reduces the per-call cost by ninety percent compared with the usual rate charged by the gateway. Because the reduction is tied to the provider order, any request that does not reach Novita will be billed at the regular price, so engineers should monitor fallback occurrences. The limited-time nature of the offer means teams should evaluate whether the cost savings justify any additional routing logic before the deadline.

The fallback mechanism is built into the gateway: if Novita is unavailable, the request automatically routes to alternative providers without error, but at the standard fee. After August 11, the special pricing disappears and the model continues to be offered at its regular rate, with no special ordering required. Consequently, any long-term cost planning must assume the post-discount pricing once the promotion ends.

Implementing the discount requires creating an API key in the AI Gateway section of the Vercel dashboard and optionally testing the model in the playground first. The code snippet shows the use of the `streamText` helper with the model name and the `gateway.order` option set to `['novita']`. Engineers should ensure their deployment pipelines include this configuration to capture the discount consistently across environments.