Be part of our each day and weekly newsletters for the most recent updates and unique content material on industry-leading AI protection. Be taught Extra
Your entire AI panorama shifted again in January 2025 after a then little-known Chinese language AI startup DeepSeek (a subsidiary of the Hong Kong-based quantitative evaluation agency Excessive-Flyer Capital Administration) launched its highly effective open supply language reasoning mannequin DeepSeek R1 publicly to the world, besting the efficiency of U.S. tech giants corresponding to Meta.
As DeepSeek utilization unfold quickly amongst researchers and enterprises, Meta was reportedly despatched into panic mode upon studying that this new R1 mannequin had been educated for a fraction of the price of many different main fashions, as little as a number of million {dollars} — what it pays a few of its personal AI crew leaders — but nonetheless achieved prime efficiency within the open supply class.
Meta’s complete generative AI technique as much as that time had been predicated on releasing best-in-class open supply fashions below its model identify “Llama” for researchers and firms to construct upon freely (no less than, if that they had fewer than 700 million month-to-month customers, at which level they’re presupposed to contact Meta for particular paid licensing phrases).
But DeepSeek R1’s astonishingly good efficiency on a much smaller finances had allegedly shaken the corporate management and compelled some form of reckoning, with the final model of Llama, 3.3, having been launched only a month prior in December 2024 but already wanting outdated.
Now we all know the fruits of that reckoning: right now, Meta founder and CEO Mark Zuckerberg took to his Instagram account to announce a new Llama 4 collection of fashions, with two of them — the 400-billion parameter Llama 4 Maverick and 109-billion parameter Llama 4 Scout — out there right now for builders to obtain and start utilizing or fine-tuning now on llama.com and AI code sharing group Hugging Face.
An enormous 2-trillion parameter Llama 4 Behemoth can be being previewed right now, although Meta’s weblog publish on the releases mentioned it was nonetheless being educated, and gave no indication of when it may be launched. (Recall parameters seek advice from the settings that govern the mannequin’s habits and that typically extra imply a extra highly effective and sophisticated throughout mannequin.)
One headline characteristic of those fashions is that they’re all multimodal — educated on, and due to this fact, able to receiving and producing textual content, video, and imagery (audio was not talked about).
One other is that they’ve extremely lengthy context home windows — 1 million tokens for Llama 4 Maverick and 10 million for Llama 4 Scout — which is equal to about 1,500 and 15,000 pages of textual content, respectively, all of which the mannequin can deal with in a single enter/output interplay. Which means a person might theoretically add or paste as much as 7,500 pages-worth-of textual content and obtain that a lot in return from Llama 4 Scout, which might be useful for information-dense fields corresponding to drugs, science, engineering, arithmetic, literature and so on.
Right here’s what else we’ve realized about this launch to date:
All-in on mixture-of-experts
All three fashions use the “mixture-of-experts (MoE)” structure method popularized in earlier mannequin releases from OpenAI and Mistral, which primarily combines a number of smaller fashions specialised (“consultants”) in several duties, topics and media codecs right into a unified complete, bigger mannequin. Every Llama 4 launch is claimed to be due to this fact a combination of 128 completely different consultants, and extra environment friendly to run as a result of solely the professional wanted for a specific activity, plus a “shared” professional, handles every token, as an alternative of the complete mannequin having to run for every one.
Because the Llama 4 weblog publish notes:
Consequently, whereas all parameters are saved in reminiscence, solely a subset of the full parameters are activated whereas serving these fashions. This improves inference effectivity by reducing mannequin serving prices and latency—Llama 4 Maverick could be run on a single [Nvidia] H100 DGX host for simple deployment, or with distributed inference for max effectivity.
Each Scout and Maverick can be found to the general public for self-hosting, whereas no hosted API or pricing tiers have been introduced for official Meta infrastructure. As a substitute, Meta focuses on distribution by open obtain and integration with Meta AI in WhatsApp, Messenger, Instagram, and internet.
Meta estimates the inference value for Llama 4 Maverick at $0.19 to $0.49 per 1 million tokens (utilizing a 3:1 mix of enter and output). This makes it considerably cheaper than proprietary fashions like GPT-4o, which is estimated to value $4.38 per million tokens, based mostly on group benchmarks.
Certainly, shortly after this publish was revealed, I acquired phrase that cloud AI inference supplier Groq has enabled Llama 4 Scout and Maverick on the following costs:
- Llama 4 Scout: $0.11 / M enter tokens and $0.34 / M output tokens, at a blended price of $0.13
- Llama 4 Maverick: $0.50 / M enter tokens and $0.77 / M output tokens, at a blended price of $0.53
All three Llama 4 fashions—particularly Maverick and Behemoth—are explicitly designed for reasoning, coding, and step-by-step downside fixing — although they don’t seem to exhibit the chains-of-thought of devoted reasoning fashions such because the OpenAI “o” collection, nor DeepSeek R1.
As a substitute, they appear designed to compete extra straight with “classical,” non-reasoning LLMs and multimodal fashions corresponding to OpenAI’s GPT-4o and DeepSeek’s V3 — aside from Llama 4 Behemoth, which does seem to threaten DeepSeek R1 (extra on this beneath!)
As well as, for Llama 4, Meta constructed customized post-training pipelines targeted on enhancing reasoning, corresponding to:
- Eradicating over 50% of “simple” prompts throughout supervised fine-tuning.
- Adopting a steady reinforcement studying loop with progressively more durable prompts.
- Utilizing cross@ok analysis and curriculum sampling to strengthen efficiency in math, logic, and coding.
- Implementing MetaP, a brand new method that lets engineers tune hyperparameters (like per-layer studying charges) on fashions and apply them to different mannequin sizes and kinds of tokens whereas preserving the supposed mannequin habits.
MetaP is of explicit curiosity because it may very well be used going ahead to set hyperparameters on on mannequin after which get many different kinds of fashions out of it, growing coaching effectivity.
As my VentureBeat colleague and LLM professional Ben Dickson opined ont the brand new MetaP method: “This may save a number of money and time. It signifies that they run experiments on the smaller fashions as an alternative of doing them on the large-scale ones.”
That is particularly vital when coaching fashions as massive as Behemoth, which makes use of 32K GPUs and FP8 precision, attaining 390 TFLOPs/GPU over greater than 30 trillion tokens—greater than double the Llama 3 coaching information.
In different phrases: the researchers can inform the mannequin broadly how they need it to behave, and apply this to bigger and smaller model of the mannequin, and throughout completely different types of media.
A strong – however not but the most highly effective — mannequin household
In his announcement video on Instagram (a Meta subsidiary, naturally), Meta CEO Mark Zuckerberg mentioned that the corporate’s “purpose is to construct the world’s main AI, open supply it, and make it universally accessible so that everybody on this planet advantages…I’ve mentioned for some time that I believe open supply AI goes to grow to be the main fashions, and with Llama 4, that’s beginning to occur.”
It’s a clearly rigorously worded assertion, as is Meta’s weblog publish calling Llama 4 Scout, “the perfect multimodal mannequin on this planet in its class and is extra highly effective than all earlier technology Llama fashions,” (emphasis added by me).
In different phrases, these are very highly effective fashions, close to the highest of the heap in comparison with others of their parameter-size class, however not essentially setting new efficiency data. Nonetheless, Meta was eager to trumpet the fashions its new Llama 4 household beats, amongst them:
Llama 4 Behemoth
- Outperforms GPT-4.5, Gemini 2.0 Professional, and Claude Sonnet 3.7 on:
- MATH-500 (95.0)
- GPQA Diamond (73.7)
- MMLU Professional (82.2)
Llama 4 Maverick
- Beats GPT-4o and Gemini 2.0 Flash on most multimodal reasoning benchmarks:
- ChartQA, DocVQA, MathVista, MMMU
- Aggressive with DeepSeek v3.1 (45.8B params) whereas utilizing lower than half the energetic parameters (17B)
- Benchmark scores:
- ChartQA: 90.0 (vs. GPT-4o’s 85.7)
- DocVQA: 94.4 (vs. 92.8)
- MMLU Professional: 80.5
- Price-effective: $0.19–$0.49 per 1M tokens
Llama 4 Scout
- Matches or outperforms fashions like Mistral 3.1, Gemini 2.0 Flash-Lite, and Gemma 3 on:
- DocVQA: 94.4
- MMLU Professional: 74.3
- MathVista: 70.7
- Unmatched 10M token context size—perfect for lengthy paperwork, codebases, or multi-turn evaluation
- Designed for environment friendly deployment on a single H100 GPU
However in spite of everything that, how does Llama 4 stack as much as DeepSeek?
However in fact, there are an entire different class of reasoning-heavy fashions corresponding to DeepSeek R1, OpenAI’s “o” collection (like GPT-4o), Gemini 2.0, and Claude Sonnet.
Utilizing the highest-parameter mannequin benchmarked—Llama 4 Behemoth—and evaluating it to the intial DeepSeek R1 launch chart for R1-32B and OpenAI o1 fashions, right here’s how Llama 4 Behemoth stacks up:
| Benchmark | Llama 4 Behemoth | DeepSeek R1 | OpenAI o1-1217 |
|---|---|---|---|
| MATH-500 | 95.0 | 97.3 | 96.4 |
| GPQA Diamond | 73.7 | 71.5 | 75.7 |
| MMLU | 82.2 | 90.8 | 91.8 |
What can we conclude?
- MATH-500: Llama 4 Behemoth is barely behind DeepSeek R1 and OpenAI o1.
- GPQA Diamond: Behemoth is forward of DeepSeek R1, however behind OpenAI o1.
- MMLU: Behemoth trails each, however nonetheless outperforms Gemini 2.0 Professional and GPT-4.5.
Takeaway: Whereas DeepSeek R1 and OpenAI o1 edge out Behemoth on a pair metrics, Llama 4 Behemoth stays extremely aggressive and performs at or close to the highest of the reasoning leaderboard in its class.
Security and fewer political ‘bias’
Meta additionally emphasised mannequin alignment and security by introducing instruments like Llama Guard, Immediate Guard, and CyberSecEval to assist builders detect unsafe enter/output or adversarial prompts, and implementing Generative Offensive Agent Testing (GOAT) for automated red-teaming.
The corporate additionally claims Llama 4 exhibits substantial enchancment on “political bias” and says “particularly, [leading LLMs] traditionally have leaned left in the case of debated political and social matters,” that that Llama 4 does higher at courting the precise wing…in line with Zuckerberg’s embrace of Republican U.S. president Donald J. Trump and his get together following the 2024 election.
The place Llama 4 stands to date
Meta’s Llama 4 fashions carry collectively effectivity, openness, and high-end efficiency throughout multimodal and reasoning duties.
With Scout and Maverick now publicly out there and Behemoth previewed as a state-of-the-art trainer mannequin, the Llama ecosystem is positioned to supply a aggressive open different to top-tier proprietary fashions from OpenAI, Anthropic, DeepSeek, and Google.
Whether or not you’re constructing enterprise-scale assistants, AI analysis pipelines, or long-context analytical instruments, Llama 4 presents versatile, high-performance choices with a transparent orientation towards reasoning-first design.
