Fortytwo (@fortytwo) header
Fortytwo (@fortytwo) profile picture

The swarm inference company

Mountain View, CA
Joined April 2024
Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)Photo by Fortytwo (@fortytwo)
Fortytwo retweeted
We built Fortytwo Engine to compete on intelligence per dollar, but most inference providers we see compete only on price and speed, without revealing how much accuracy gets lost in those optimizations. Thanks @fluence_project for inviting us to write about these issues
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇
4
8
1
36
3,041
1b+ tokens were processed on Fortytwo Engine this week
10
10
3
60
171,514
Fortytwo retweeted
Ivan Nikitin (@inikitin) co-founder and CEO at @fortytwo says the AI race doesn't have to be won by whoever controls the biggest data center, chips, and energy supply. Small specialized models are far more efficient and Fortytwo connects many of them into one network. Working together they can beat frontier models while running on ordinary laptops.
6
6
24
1,810
Fortytwo retweeted
Two great days at @hub71ad Impact Defined 2026 in Abu Dhabi. We showcased Swarm Inference at the Fortytwo booth and pitched it from the stage. The strongest conversations were around sovereign AI: how enterprises and governments can get frontier-grade intelligence where their data already lives, using specialized models and swarm infrastructure instead of relying entirely on centralized AI providers.
9
11
56
2,094
Fortytwo retweeted
Decentralized AI is our industry's response to increasingly centralized SuperLabs
24
25
1
258
12,034
Fortytwo retweeted
Fortytwo's first product, Fortytwo Prime, allows agents to get the best answer from a swarm of frontier models. Decentralize Intelligence.
Fortytwo Prime runs a curated swarm of frontier models in parallel and returns the strongest answer to your agent via MCP Payments settle onchain on Monad and Base: agents pay only for tokens consumed We extended x402 with an escrow layer to make this work ↓
18
23
1
184
7,481
Fortytwo retweeted
Where AI Agents Meet DePIN. The next wave of Web3 is here. Join us for a conversation with @Acurast × @fortytwo × @SecretNetwork Exploring the intersection of AI agents, decentralized compute, and what’s coming next. Compute. Privacy. Ownership. March 19, 17:00 UTC Set your reminder 👇 teapaw.org/i/spaces/1qJVmQZEnXdGB…
94
47
6
124
18,165
Meet Strand-Rust-Coder-14B, a specialized AI that writes Rust code like a senior developer. It's not just another coding assistant, it's specifically fine-tuned for Rust, making it a game-changer for systems programming and performance-critical applications. This is exactly what the Rust community has been waiting for.
47
128
30
1,171
92,931
Compute is the most competitive sector, with 50+ projects and relatively little differentiation.
2
4
3
48
7,106
AI is hitting a scalability wall — centralized compute is capped, training costs are exploding, and frontier models are showing diminishing returns. A new architecture is emerging: decentralized compute + swarm intelligence 🧠 In our new episode, @TheTomTrow talks with @inikitin — co-founder and CEO @fortytwo about how small specialized models, peer ranking and a global node network are already outperforming GPT-5–class systems on real benchmarks 🎧 Listen on all platforms — this one is a must for anyone watching the future of AI infrastructure.
10
38
3
185
5,081
🚨State of AI special episode of Fully Diluted 🚨 @capradavis sits down with @inikitin co-founder of @fortytwo to discuss: -- the benefits of small, specialized models over one "god model" -- the hard walls of AI scalability
21
15
8
75
26,863
The road to AGI. The founders of @fortytwo lay out why they have a big breakthrough with its decentralized AI that will get to AGI faster than the "big" LLM companies. Decentralized "swarms" of AIs working together is a big idea and will lead us to faster learning, lighter AIs that can run on small computers at the edge. Thanks @inikitin and @vlarin for such an interesting conversation. In comments I'll post what Grok learned by watching this.
27
32
11
202
35,463
Swarm inference is a brilliantly powerful idea. 💡 Turns a pile of small models into one frontier-level answer that beats most SOTA single-model setups. In this paper @fortytwo wires many small models into a peer-ranked swarm that answers as one, and it beats big single SOTA models by using head-to-head judging plus reputation weighting. Reports an incredible 85.90% on GPQA Diamond, 99.6% on MATH-500, 100% on AIME 2024, 96.66% on AIME 2025, and 84.4% on LiveCodeBench, with the swarm beating the same models under majority voting by +17.21 points. The core move is pairwise ranking with a Bradley‑Terry style aggregator. Each question fans out to many nodes where various expert models generate their responses, so the swarm has lots of different candidates to compare. Each node runs an SLM (Small Language Model), produces its own answer, and the swarm behaves like a single model to the outside user. The same nodes both generate answers and judge them, they compare pairs head to head and add a tiny rationale for which answer is better. A Bradley-Terry style aggregator turns all those wins and losses into a single stable ranking, which consistently beats simple majority voting because it uses the strength of every matchup, not just raw counts. With this paper, Fortytwo brings a new approach in decentralized artificial intelligence, demonstrating that collective intelligence can exceed individual capabilities while maintaining practical deployability. 🧵 Read on 👇
9
39
3
188
14,545
When I invested in FortyTwo, the question wasn’t whether scaling would end. It was whether scaling alone could account for intelligence. The prevailing paradigm has treated cognition as a function of magnitude: enlarge the transformer, deepen the context window, harvest more data. What Forty-Two explored instead was intelligence as a property of coordinated systems rather than isolated models. Melanie Mitchell, in Complexity: A Guided Tour, describes complex adaptive systems as those “whose collective behavior cannot be predicted from the behavior of their parts.” That observation provides a useful frame for what’s now unfolding in AI. Forty-Two’s network is composed of heterogeneous agents, retrievers, reasoners, evaluators linked by feedback and self-correction loops. Each module is limited in scope; together they exhibit competence that isn’t linearly attributable to any component. The recent benchmarks prove this out. Performance gains did not arise from additional parameters but from the coupling of subsystems from interaction terms rather than scale factors. The architecture behaves less like a monolithic brain and more like a distributed ecology, where meaning and inference emerge from information flow. “Computation is a natural process; nature computes itself.” Forty-Two’s result suggests that intelligence, too, may be an emergent property of computation in motion, an outcome of ongoing coordination rather than a static accumulation of weights.
Announcing Fortytwo’s Swarm Inference A decentralized AI architecture that outperforms the top frontier models from the biggest labs: > ChatGPT 5 (OpenAI), > Gemini 2.5 Pro (Google), > Claude Opus 4.1 (Anthropic), > Grok 4 (xAI), > DeepSeek R1 (DeepSeek). Thread ↓
6
7
83
8,593
Swarm Inference is a game changer for the AI industry. To understand why, you need to start with what the returns look like for the builders of regular inference suppliers (GPUs in data centers). Put on your investor hat: -You see the boom of AI. You want to get involved. Buying GPUs, putting them somewhere, and renting them out is the core of this business model. - You start modeling, you build your PnL. Data centers, servers, connectivity, etc. This is your CAPEX and COGS. Top-line stands out only for high-end GPUs. Power is expensive. Management isn't trivial. - You arrive at a quick conclusion: the returns are... rough to say the least. If you run through this exercise, there's a simple TLDR: Contributing to the AI economy at the infrastructure level only works with scale, or some connect on power/space. Inference infrastructure is commoditized and yield has compressed. I keep trying to find a way to slice and dice this, but it just feels like there's simply no alpha to be found here. From a technology perspective, this is all a consequence of the way Inference has historically worked today. You need compute density. High bus throughput, high VRAM, etc. Swarm Inference is a game changer because compute density is no longer a requirement. The CAPEX and COGS above are a consequence of the necessity of compute density. But if you don't need compute density, then you don't need data centers. Said differently, individual infrastructure providers don't have to pass a scaling hurdle to participate. And if you don't need minimum scale to participate, then maybe even an individual home can contribute to AI with returns that actually make sense. That's why Swarm Inference is a game changer. It's innovation that changes where alpha can be found. Congrats to the team at Fortytwo for executing on a paradigm shift.
Announcing Fortytwo’s Swarm Inference A decentralized AI architecture that outperforms the top frontier models from the biggest labs: > ChatGPT 5 (OpenAI), > Gemini 2.5 Pro (Google), > Claude Opus 4.1 (Anthropic), > Grok 4 (xAI), > DeepSeek R1 (DeepSeek). Thread ↓
12
8
79
48,654
5 < 42
79
38
15
330
47,843
Fortytwo retweeted
Compared @fortytwo swarm inference to OpenAI: - Centralized reasoning LLMs tend to degrade on complex tasks. - Fortytwo’s reranking-based swarm inference works — 3× faster, 10× cheaper, and gets the job done.
8
6
6
56
27,500
ok guys i just completed a cultural study
17
11
3
106
7,802
Fortytwo retweeted
At @fortytwo ‘s Thinking Machines Day with @MRRydon @inikitin @Ramkumartweet @wagmigently
96
45
1
518
18,461