vs
xAI vs Cohere
xAI's Grok pairs cheap output tokens with live X data. Cohere pairs enterprise retrieval models with deployment inside a customer's own network.
By The Subconscious Team · Updated
xAI vs Cohere: key differences
Both are labs selling their own models, aimed at very different buyers. Grok 4.6 has a 500K context window at $2 in and $6 out under 200K prompt tokens, and the older Grok 4.20 and 4.3 keep a 1M window at $1.25 in and $2.50 out. Past 200K tokens the whole request bills at double. Cohere's Command A costs $2.50 in and $10 out with 256K context, Command A+ runs 128K, and Command R7B starts at $0.0375 in. On output price and context length, xAI is ahead. Grok's distinct feature is server-side Web Search and X Search for live posts and news.
Cohere wins on enterprise fit. xAI serves only its first-party API and has a smaller enterprise and cloud-marketplace footprint. Cohere runs on its API, Bedrock, SageMaker, Azure AI Foundry and OCI, and supports private VPC or on-prem deployment with fine-tuning. Command A+ weights are open under Apache 2.0, while Grok is closed. Embed 4 and Rerank 4 add retrieval xAI does not offer. xAI does ship separate first-party APIs for image, video and audio, a breadth Cohere matches only partly with Transcribe. For social listening and real-time market agents, Grok. For grounded answers over internal documents under strict data rules, Cohere.
What xAI and Cohere 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 profileCohere
Cohere is a Toronto-based lab that sells models and platforms to banks, governments and large enterprises rather than consumers. Its generative line is the Command family. Command A+, released May 20, 2026, is a 218B-parameter mixture-of-experts model with 25B active, published under Apache 2.0 with a 128K context window, and it combines reasoning, vision, translation and tool use in one set of weights. Command A has a 256K window and lists at $2.50 in and $10 out per million tokens, while Command R7B costs $0.0375 in. June 2026 added North Mini Code, a 30B Apache 2.0 coding model, and the lineup also includes Aya multilingual models and Transcribe for speech.
Example models: Command A+, Command A, Embed 4, Rerank 4
Full Cohere profileShould you choose xAI or Cohere?
xAI vs Cohere at a glance
| Attribute | xAI | Cohere |
|---|---|---|
| Model access | Closed | Closed, plus open Command A+ |
| Flagship models | Grok 4.6, Grok 4.20, grok-build | Command A+, Command A, Embed 4, Rerank 4 |
| Speed | ~54 tok/s on Grok 4.6 | 375 tok/s on Command A+ W4A4, per Cohere |
| Price | $2 in, $6 out (Grok 4.6); 2x past 200K | $0.0375–$2.50 in, $0.15–$10 out per 1M |
| Customization | Unknown | Enterprise fine-tuning, incl. private |
| Deployment | First-party API | API, Bedrock, Azure, OCI, VPC, on-prem |
| Long context | 500K (4.6), 1M (4.20, 4.3) | 256K on Command A; 128K on A+ |
Frequently asked questions
What is the difference between xAI and Cohere?
xAI's Grok pairs cheap output tokens with live X data. Cohere pairs enterprise retrieval models with deployment inside a customer's own network.
When should I choose xAI over Cohere?
Agents that need live X and web data; Cheap output tokens on reasoning work; Up to 1M context on Grok 4.20.
When should I choose Cohere over xAI?
Internal document search in a private cloud; Open weights to self-host under Apache 2.0; Buying through Bedrock, Azure or OCI.
Is xAI or Cohere cheaper?
xAI: $2 in, $6 out (Grok 4.6); 2x past 200K. Cohere: $0.0375–$2.50 in, $0.15–$10 out per 1M. The cheaper choice depends on the model and workload.
Which has more context, xAI or Cohere?
xAI: 500K (4.6), 1M (4.20, 4.3). Cohere: 256K on Command A; 128K on A+.
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.