Kimi K3, First Open 3T-Class Model, Lands on Telnyx Inference API
Moonshot AI's 2.8-trillion-parameter model with 1M-token context is now accessible via Telnyx's owned GPU infrastructure at $2.70 per million input tokens.
Telnyx has added Kimi K3 to its Inference API, giving developers access to the largest open-weight language model ever released. The 2.8-trillion-parameter model from Moonshot AI became available on July 27, 2026, when its weights were published on Hugging Face.
Model Specifications
Kimi K3 is the first open-source model to reach the 3-trillion-parameter class. It features a 1-million-token context window and native multimodal capabilities spanning text, images, and video. The API supports configurable reasoning effort levels (low, high, max), tool calling, and automatic prompt caching.
Pricing and Infrastructure
Telnyx is pricing Kimi K3 at $2.70 per million input tokens, $13.50 per million output tokens, and $0.27 per million cached input tokens. The model runs on Telnyx's own GPU infrastructure deployed across the US, EU, APAC, and MENA regions, rather than on rented cloud capacity.
The company positions this owned infrastructure as a competitive advantage. "The competitive advantage in AI is shifting from who builds the smartest model to who builds the infrastructure that decides where every request runs," Telnyx said in its release notes.
Benchmark Performance
According to Moonshot's own benchmarks, Kimi K3 trails Claude Fable 5 and GPT 5.6 Sol on aggregate benchmarks for coding and agentic work. Independent third-party verification of this ranking is limited.
Developer Access
The model is accessible through an OpenAI-compatible API endpoint, allowing developers to integrate Kimi K3 into existing workflows without modifying their code. The standardized interface lowers the barrier to deploying high-reasoning, long-context agentic workflows without managing massive GPU clusters.
Moonshot AI has been aggressively scaling its Kimi series, moving from K2.6 to K3. By releasing open weights for a model of this scale, Moonshot is challenging the dominance of closed-source frontier models from OpenAI and Anthropic.
Telnyx's integration signals a broader shift in the AI infrastructure layer, where providers compete on latency, cost, and geographic coverage rather than model development alone.