Kimi K3, DeepSeek and GLM claim to be free of NVIDIA – testing reveals the catch
deepseek nvidia
| Source: Mastodon | Original article
NVIDIA Build now provides free access to Kimi K3, DeepSeek V4.1 Flash and GLM 5.3 without daily caps, though users must verify via phone and availability varies by country.
Nvidia has rolled out a free‑tier API program that gives developers unrestricted access to more than 80 large‑language models, including Kimi K3, GLM 5.3 and DeepSeek V4.1 Flash. The “Nvidia Build” offering removes the daily‑usage caps that typically accompany trial keys, allowing unlimited calls to these models for coding, agent‑building and other development tasks.
The promise of open, cost‑free access is tempered by two practical hurdles. First, an SMS‑based phone verification is required before an API key is issued, a step that adds friction for hobbyists and small teams. Second, the service is not available in every country; the rollout currently excludes a handful of regions, limiting the program’s inclusivity. Those restrictions were confirmed during a hands‑on test of the claim, which found the models reachable once the phone check was completed, but inaccessible from several locations.
Why it matters is twofold. For the burgeoning open‑source AI community, free, uncapped endpoints lower the barrier to experimenting with state‑of‑the‑art models that would otherwise demand expensive cloud credits or on‑prem hardware. At the same time, the verification and geo‑limits preserve a degree of control for Nvidia, hinting that the company is balancing openness with regulatory and commercial considerations.
Looking ahead, developers will be watching whether Nvidia expands the free tier to cover more territories or relaxes the phone‑check requirement. Competitors may respond with their own unrestricted APIs, and the broader ecosystem could see a shift toward wider adoption of open‑weight models such as DeepSeek’s MIT‑licensed V4.1 and GLM’s Chinese‑backed 5.3. The next update from Nvidia could determine how quickly these models move from niche tools to mainstream building blocks for AI‑driven products.
Sources
Back to AIPULSEN