# Thinking Machines vs Particle.AI

> Particle.AI serves cheap Flash-class open models with 1M context through Vercel AI Gateway. Thinking Machines trains open models and serves only its larger Inkling models in beta.

Canonical: https://www.subconscious.dev/compare/thinking-machines-vs-particle-ai · By The Subconscious Team · Updated September 30, 2026

## How they compare

Both offer 1M context, at very different price points. Particle lists GLM 5.3 Flash at $0.10 in and $0.40 out, DeepSeek V4 Flash 0731 at $0.14 in and $0.28 out, and DeepSeek V4.1 Flash at $0.25 in and $1 out with about 157 tokens per second, all with cache reads at $0.03 per million. Thinking Machines' Inkling costs $1.00 in and $4.05 out, but it is a 975B MoE with 41B active and adds image and audio input. Particle has no customization listed. Thinking Machines' main product is Tinker, which runs custom LoRA SFT and RL on bases like GLM-5.3 and Kimi K2.6.

Access differs too. Particle is reachable through Vercel AI Gateway with no new contract, which makes it easy as a cheap fallback route. It is a very early company with a tiny catalog, and some listings trail faster hosts, like 3.5 seconds of latency on DeepSeek V4.1 Flash. Thinking Machines is far better funded, but its serverless API is beta and its checkpoint endpoint is meant for testing and low internal traffic. For cheap high-volume text calls, Particle wins. For training a specialized model or testing native audio input, Thinking Machines is the one to use.

## What each one does

### Thinking Machines

Thinking Machines Lab is the San Francisco lab Mira Murati, formerly CTO of OpenAI, founded in February 2025, with OpenAI co-founder John Schulman as chief scientist. It raised about $2 billion at a $12 billion valuation in July 2025 in a round led by Andreessen Horowitz, and in March 2026 signed a multi-year Nvidia deal for one gigawatt of Vera Rubin capacity. Its main developer product is Tinker, an API for post-training open-weight models that launched in October 2025 and is now generally available. Tinker exposes four low-level calls, forward_backward, optim_step, sample and save_state, so teams write their own supervised or reinforcement learning loops while Thinking Machines runs the distributed GPU work. Training uses LoRA adapters rather than full weight updates.

### Particle.AI

Particle AI is an early San Francisco infrastructure startup with a mission to make intelligence as cheap and abundant as electricity. The team works on post-training, inference optimization and distributed systems, all aimed at pushing down cost per unit of intelligence. It is still hiring its founding team and works fully in person. Public detail about funding and founders is thin as of this writing.

## Which is best, and when

### Choose Thinking Machines for

- LoRA SFT and RL on open weights
- Image and audio input on a large MoE
- Research on custom post-training

### Choose Particle.AI for

- Cheap high-volume Flash model calls
- A price-optimized gateway fallback
- 1M context at low per-token rates

## At a glance

| Attribute | Thinking Machines | Particle.AI |
|---|---|---|
| Model access | Open weights | Open weights |
| Flagship models | Inkling, Inkling-Small | DeepSeek V4.1 Flash, GLM 5.3 Flash |
| Speed | - | ~157 tok/s on DeepSeek V4.1 Flash |
| Price | Per 1M tokens by prefill, sample, train; Inkling $1.00 in, $4.05 out | $0.10 in, $0.40 out (GLM 5.3 Flash) |
| Customization | LoRA SFT and RL via Tinker | - |
| Deployment | Training API, beta serverless (Inkling only) | Via Vercel AI Gateway |
| Long context | Inkling up to 1M; Tinker 32K–256K | 1M |

## FAQ

### What is the difference between Thinking Machines and Particle.AI?

Particle.AI serves cheap Flash-class open models with 1M context through Vercel AI Gateway. Thinking Machines trains open models and serves only its larger Inkling models in beta.

### When should I choose Thinking Machines over Particle.AI?

LoRA SFT and RL on open weights; Image and audio input on a large MoE; Research on custom post-training.

### When should I choose Particle.AI over Thinking Machines?

Cheap high-volume Flash model calls; A price-optimized gateway fallback; 1M context at low per-token rates.

### Is Thinking Machines or Particle.AI cheaper?

Thinking Machines: Per 1M tokens by prefill, sample, train; Inkling $1.00 in, $4.05 out. Particle.AI: $0.10 in, $0.40 out (GLM 5.3 Flash). The cheaper choice depends on the model and workload.

### Which has more context, Thinking Machines or Particle.AI?

Thinking Machines: Inkling up to 1M; Tinker 32K–256K. Particle.AI: 1M.

## Running long-horizon agents?

If your agents run past 200K tokens, compare both against Subconscious: [Subconscious vs Thinking Machines](https://www.subconscious.dev/compare/subconscious-vs-thinking-machines.md), [Subconscious vs Particle.AI](https://www.subconscious.dev/compare/subconscious-vs-particle-ai.md).

Full profiles: [Thinking Machines](https://www.subconscious.dev/providers/thinking-machines.md), [Particle.AI](https://www.subconscious.dev/providers/particle-ai.md).