Be a part of our each day and weekly newsletters for the newest updates and unique content material on industry-leading AI protection. Study Extra
Meta introduced at the moment a partnership with Cerebras Techniques to energy its new Llama API, providing builders entry to inference speeds as much as 18 occasions quicker than conventional GPU-based options.
The announcement, made at Meta’s inaugural LlamaCon developer convention in Menlo Park, positions the corporate to compete straight with OpenAI, Anthropic, and Google within the quickly rising AI inference service market, the place builders buy tokens by the billions to energy their purposes.
“Meta has chosen Cerebras to collaborate to ship the ultra-fast inference that they should serve builders via their new Llama API,” stated Julie Shin Choi, chief advertising and marketing officer at Cerebras, throughout a press briefing. “We at Cerebras are actually, actually excited to announce our first CSP hyperscaler partnership to ship ultra-fast inference to all builders.”
The partnership marks Meta’s formal entry into the enterprise of promoting AI computation, reworking its common open-source Llama fashions right into a business service. Whereas Meta’s Llama fashions have gathered over one billion downloads, till now the corporate had not provided a first-party cloud infrastructure for builders to construct purposes with them.
“That is very thrilling, even with out speaking about Cerebras particularly,” stated James Wang, a senior govt at Cerebras. “OpenAI, Anthropic, Google — they’ve constructed a complete new AI enterprise from scratch, which is the AI inference enterprise. Builders who’re constructing AI apps will purchase tokens by the thousands and thousands, by the billions typically. And these are similar to the brand new compute directions that individuals must construct AI purposes.”
Breaking the pace barrier: How Cerebras supercharges Llama fashions
What units Meta’s providing aside is the dramatic pace enhance supplied by Cerebras’ specialised AI chips. The Cerebras system delivers over 2,600 tokens per second for Llama 4 Scout, in comparison with roughly 130 tokens per second for ChatGPT and round 25 tokens per second for DeepSeek, in accordance with benchmarks from Synthetic Evaluation.
“In the event you simply evaluate on API-to-API foundation, Gemini and GPT, they’re all nice fashions, however all of them run at GPU speeds, which is roughly about 100 tokens per second,” Wang defined. “And 100 tokens per second is okay for chat, however it’s very sluggish for reasoning. It’s very sluggish for brokers. And persons are combating that at the moment.”
This pace benefit allows solely new classes of purposes that had been beforehand impractical, together with real-time brokers, conversational low-latency voice methods, interactive code era, and immediate multi-step reasoning — all of which require chaining a number of massive language mannequin calls that may now be accomplished in seconds somewhat than minutes.
The Llama API represents a major shift in Meta’s AI technique, transitioning from primarily being a mannequin supplier to turning into a full-service AI infrastructure firm. By providing an API service, Meta is making a income stream from its AI investments whereas sustaining its dedication to open fashions.
“Meta is now within the enterprise of promoting tokens, and it’s nice for the American type of AI ecosystem,” Wang famous throughout the press convention. “They carry lots to the desk.”
The API will supply instruments for fine-tuning and analysis, beginning with Llama 3.3 8B mannequin, permitting builders to generate knowledge, prepare on it, and take a look at the standard of their customized fashions. Meta emphasizes that it received’t use buyer knowledge to coach its personal fashions, and fashions constructed utilizing the Llama API will be transferred to different hosts—a transparent differentiation from some rivals’ extra closed approaches.
Cerebras will energy Meta’s new service via its community of knowledge facilities situated all through North America, together with services in Dallas, Oklahoma, Minnesota, Montreal, and California.
“All of our knowledge facilities that serve inference are in North America presently,” Choi defined. “We will likely be serving Meta with the total capability of Cerebras. The workload will likely be balanced throughout all of those totally different knowledge facilities.”
The enterprise association follows what Choi described as “the traditional compute supplier to a hyperscaler” mannequin, just like how Nvidia supplies {hardware} to main cloud suppliers. “They’re reserving blocks of our compute that they will serve their developer inhabitants,” she stated.
Past Cerebras, Meta has additionally introduced a partnership with Groq to offer quick inference choices, giving builders a number of high-performance options past conventional GPU-based inference.
Meta’s entry into the inference API market with superior efficiency metrics may probably disrupt the established order dominated by OpenAI, Google, and Anthropic. By combining the recognition of its open-source fashions with dramatically quicker inference capabilities, Meta is positioning itself as a formidable competitor within the business AI house.
“Meta is in a singular place with 3 billion customers, hyper-scale datacenters, and an enormous developer ecosystem,” in accordance with Cerebras’ presentation supplies. The mixing of Cerebras expertise “helps Meta leapfrog OpenAI and Google in efficiency by roughly 20x.”
For Cerebras, this partnership represents a significant milestone and validation of its specialised AI {hardware} method. “Now we have been constructing this wafer-scale engine for years, and we at all times knew that the expertise’s first fee, however finally it has to finish up as a part of another person’s hyperscale cloud. That was the ultimate goal from a business technique perspective, and we’ve lastly reached that milestone,” Wang stated.
The Llama API is presently obtainable as a restricted preview, with Meta planning a broader rollout within the coming weeks and months. Builders enthusiastic about accessing the ultra-fast Llama 4 inference can request early entry by choosing Cerebras from the mannequin choices inside the Llama API.
“In the event you think about a developer who doesn’t know something about Cerebras as a result of we’re a comparatively small firm, they will simply click on two buttons on Meta’s commonplace software program SDK, generate an API key, choose the Cerebras flag, after which abruptly, their tokens are being processed on an enormous wafer-scale engine,” Wang defined. “That type of having us be on the again finish of Meta’s complete developer ecosystem is simply super for us.”
Meta’s alternative of specialised silicon indicators one thing profound: within the subsequent part of AI, it’s not simply what your fashions know, however how shortly they will suppose it. In that future, pace isn’t only a function—it’s the entire level.
