Prime Inference
A large model inference service platform launched by Prime Intellect, built on vLLM and optimized for agentic workloads. It supports prefill/decode topology disaggregation, scheduler bubble optimization, and reliable tool calling.
Core Facts
When selecting an agent inference service, attention should be paid to its end-to-end performance under real agent workloads, rather than only looking at decoding speed on benchmarks.
The default models of multiple Providers have changed: Cortecs, CrossModel, Eden AI, Google Gemini, GitHub Copilot, Google Vertex AI, LLM Gateway, Ofox, OpenCode Go, and Prime Inference now default to Gemini 4 Argon; Fireworks AI, Nebius Token Factory, and Requesty switch to Muse Spark Open; CoreWeave, DigitalOcean, NanoGPT, and Nvidia adopt Nemotron 3.5 Super VL; Aixy and Neon use Claude Haiku 5.5; Agnes AI, Ambient, IteraCompute, Poe, Scaleway, and SenseNova also have their own adjustments.
Timeline (last 90 days)
Prime Intellect launched Prime Inference on October 2, 2025, and achieved production data showing the same number of GPUs carrying 75% more requests on GB200 NVL72.
Prime Inference's default model is Gemini 4 Argon, Requesty's is Muse Spark Open, Nvidia's and Nebius Token Factory's is Nemotron 3.5 Super VL, and Aixy's and Neon's is Claude Haiku 5.5.
The default models of multiple Providers have changed: Cortecs, CrossModel, Eden AI, Google Gemini, GitHub Copilot, Google Vertex AI, LLM Gateway, Ofox, OpenCode Go, and Prime Inference now default to Gemini 4 Argon; Fireworks AI, Nebius Token Factory, and Requesty switch to Muse Spark Open; CoreWeave, DigitalOcean, NanoGPT, and Nvidia adopt Nemotron 3.5 Super VL; Aixy and Neon use Claude Haiku 5.5; Agnes AI, Ambient, IteraCompute, Poe, Scaleway, and SenseNova also have their own adjustments.
Three providers, AI/ML API, LowRouter, and Prime Inference, were added to the model directory.
The v4.1.24 update refreshed the model catalog, adding Claude Haiku 5.5, as well as providers such as AI/ML API, LowRouter, and Prime Inference.
v3.0.71's model catalog adds three providers: AI/ML API, LowRouter and Prime Inference; the v0 provider is streamlined to provide only the default model `v0-1.5-md`; OpenCode Go migrates to the new endpoint `https://opencode.ai/inference/go/v1`; and Step 5 Preview is added to the Cline free list.
Prime Intellect launched Prime Inference, deploying the GLM-5.3 model at scale based on vLLM.
When selecting an agent inference service, attention should be paid to its end-to-end performance under real agent workloads, rather than only looking at decoding speed on benchmarks.
Prime Inference provides guarantees at the inference layer for structured output of tool calls and the stability of function calls.
Prime Inference incorporates specialized design in the prefill/decode topology to differentiate the allocation of compute resources between the two phases.
1 more timeline events
All Facts (10)
When selecting an agent inference service, attention should be paid to its end-to-end performance under real agent workloads, rather than only looking at decoding speed on benchmarks.
75%Partially VerifiedThe default models of multiple Providers have changed: Cortecs, CrossModel, Eden AI, Google Gemini, GitHub Copilot, Google Vertex AI, LLM Gateway, Ofox, OpenCode Go, and Prime Inference now default to Gemini 4 Argon; Fireworks AI, Nebius Token Factory, and Requesty switch to Muse Spark Open; CoreWeave, DigitalOcean, NanoGPT, and Nvidia adopt Nemotron 3.5 Super VL; Aixy and Neon use Claude Haiku 5.5; Agnes AI, Ambient, IteraCompute, Poe, Scaleway, and SenseNova also have their own adjustments.
75%UnverifiedPrime Inference's default model is Gemini 4 Argon, Requesty's is Muse Spark Open, Nvidia's and Nebius Token Factory's is Nemotron 3.5 Super VL, and Aixy's and Neon's is Claude Haiku 5.5.
50%UnverifiedThree providers, AI/ML API, LowRouter, and Prime Inference, were added to the model directory.
50%UnverifiedPrime Inference incorporates specialized design in the prefill/decode topology to differentiate the allocation of compute resources between the two phases.
50%UnverifiedPrime Inference provides guarantees at the inference layer for structured output of tool calls and the stability of function calls.
50%UnverifiedPrime Inference has been specifically optimized for agent workloads.
50%UnverifiedPrime Intellect launched Prime Inference on October 2, 2025, and achieved production data showing the same number of GPUs carrying 75% more requests on GB200 NVL72.
50%UnverifiedThe v4.1.24 update refreshed the model catalog, adding Claude Haiku 5.5, as well as providers such as AI/ML API, LowRouter, and Prime Inference.
50%UnverifiedPrime Intellect launched Prime Inference, deploying the GLM-5.3 model at scale based on vLLM.
50%