xAI vs Thinking Machines

xAI rents closed Grok models with native X search and cheap output. Thinking Machines offers open Inkling weights and a training API to specialize open models.

By The Subconscious Team · Updated September 30, 2026

xAI vs Thinking Machines: key differences

xAI serves closed Grok models through its own API. Grok 4.6 costs $2 in and $6 out under 200K prompt tokens with a 500K window, and Grok 4.20 and 4.3 keep 1M context at $1.25 in and $2.50 out. Its unique hook is server-side Web Search and X Search, which pull live posts from X. There is no customization listed. Thinking Machines comes from the opposite direction. Its own Inkling model is Apache 2.0, 975B parameters with 41B active, 1M context and native image and audio input, at $1.00 in and $4.05 out on a beta serverless API. Tinker then lets teams post-train Inkling or other open bases with custom SFT or RL loops.

On price for long prompts, xAI has a catch: once a prompt hits 200K tokens, the whole request bills at double. Inkling lists no such surcharge, though its serverless API is beta and serves only the two Inkling models. xAI is the more production-ready API and the only one with native X data, which matters for news, market and sentiment agents. Grok 4.20's output price is also lower than Inkling's. Thinking Machines is the better fit when a team wants weights it can keep, audio input, or a model trained on its own data. Neither has a large cloud-marketplace footprint.

What xAI and Thinking Machines do

xAI

xAI sells the Grok models through its own API. Grok 4.6 is the current flagship and xAI tells developers to use it for everything outside audio, image and video, code included. It has a 500K context window and costs $2 in and $6 out per million tokens under 200K prompt tokens. Older Grok 4.20 and 4.3 models keep a 1M window at $1.25 in and $2.50 out, which is aggressive for that capability class.

Example models: Grok 4.6, Grok 4.20

Full xAI profile

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.

Example models: Inkling, Inkling-Small, Qwen3.5, Kimi K2.6

Full Thinking Machines profile

Should you choose xAI or Thinking Machines?

xAI

Choose xAI for

Agents that need live X and web search
Low output prices on Grok 4.20
Production-ready closed reasoning API

Thinking Machines

Choose Thinking Machines for

Open Apache 2.0 weights instead of a closed model
Multimodal input including audio
Post-training a model on proprietary data

xAI vs Thinking Machines at a glance

Attribute	xAI	Thinking Machines
Model access	Closed	Open weights
Flagship models	Grok 4.6, Grok 4.20, grok-build	Inkling, Inkling-Small
Speed	~54 tok/s on Grok 4.6	Unknown
Price	$2 in, $6 out (Grok 4.6); 2x past 200K	Per 1M tokens by prefill, sample, train; Inkling $1.00 in, $4.05 out
Customization	Unknown	LoRA SFT and RL via Tinker
Deployment	First-party API	Training API, beta serverless (Inkling only)
Long context	500K (4.6), 1M (4.20, 4.3)	Inkling up to 1M; Tinker 32K–256K

Frequently asked questions

What is the difference between xAI and Thinking Machines?

xAI rents closed Grok models with native X search and cheap output. Thinking Machines offers open Inkling weights and a training API to specialize open models.

When should I choose xAI over Thinking Machines?

Agents that need live X and web search; Low output prices on Grok 4.20; Production-ready closed reasoning API.

When should I choose Thinking Machines over xAI?

Open Apache 2.0 weights instead of a closed model; Multimodal input including audio; Post-training a model on proprietary data.

Is xAI or Thinking Machines cheaper?

xAI: $2 in, $6 out (Grok 4.6); 2x past 200K. Thinking Machines: Per 1M tokens by prefill, sample, train; Inkling $1.00 in, $4.05 out. The cheaper choice depends on the model and workload.

Which has more context, xAI or Thinking Machines?

xAI: 500K (4.6), 1M (4.20, 4.3). Thinking Machines: Inkling up to 1M; Tinker 32K–256K.

Related comparisons

Running long-horizon agents?

If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.

Subconscious vs xAI Subconscious vs Thinking Machines