Long runs, faster and more accurate.
Open models served specifically for coding agents: 2x faster task completion, accuracy that holds past 200k tokens, and Team Token Plans with 5x the usage.
# Install and sign in
$npm install -g subconscious-cli
$subc login
# Run Claude Code in Marathon Mode
$subc claude
●Claude Code now runs on subconscious/glm-5.3-marathon
- Faster task completion
2x
- Effective context
5M+
- Higher accuracy
1-10%
- Lower cost
50%+
In the cloud, with your own key
Works with any agent that takes an OpenAI or Anthropic endpoint.
Team Token Plans with 5x the usage
Large monthly budgets for you and your team. Go past a day's usage and you move back to usage-based pricing.
Base
For teams putting agents into production
$1,000
5x usage, every day*
- Usage value
- $5,000+ / month
- Concurrent requests
- 100
- Rate limits
- Standard
TeamMost popular
For engineering teams running agents all day
$5,000
5x usage, every day*
- Usage value
- $25,000+ / month
- Concurrent requests
- 500
- Rate limits
- Increased
Heavy
For organizations running agents at scale
$10,000
5x usage, every day*
- Usage value
- $50,000+ / month
- Concurrent requests
- 1,000
- Rate limits
- Increased
* Usage resets daily at midnight UTC. Usage value is what a plan's daily usage would cost at our usage-based rates over a month, assuming a token mix of 98.2% cached tokens, 1.6% input tokens, and 0.2% output tokens, the weighting that characterizes long-horizon traces.
Built for the way agents actually use tokens.
Powerful open models, enhanced for long context reasoning.
Run longer, faster
Tasks finish 2x faster, and 5M+ effective context keeps an agent going long after a standard context window would fill.
More accurate
A cleaner context window means a cleaner view of the task: 1–10% higher accuracy on agentic benchmarks than standard inference on the same model.
No training
Prompts and completions are never used to train anyone’s model. What your agents read stays yours.
No refusals
An open model we serve ourselves, so legitimate work never trips a filter and quality never quietly degrades under load.
Point your coding agent at us and see.
Three commands and you are running on your own allocation. Same harness, same workflow, and if it is not faster on your repo you are back where you started in a minute.
What people say after their first run.
“It was faster than a lot of LLMs I've used, and it was easy to set up.”
Cedric Prentice
Software Engineer @ Wayfair
“Very cool, easy to use!”
Bill Simmons
Co-Founder @ Orbit.me
“Very interesting alternative to OpenAI”
Dave Gogi
Founder @ Signal X
“Pretty fast and fun to use”
Nihir Kothari
Founder @ Sidekick Software
“It worked very well and provided accurate descriptions of the images we passed in. The API costs were very cheap.”
Sam Mayle
Researcher @ Mitsubishi Electric Research Lab
“It was faster than a lot of LLMs I've used, and it was easy to set up.”
Cedric Prentice
Software Engineer @ Wayfair
“Very cool, easy to use!”
Bill Simmons
Co-Founder @ Orbit.me
“Very interesting alternative to OpenAI”
Dave Gogi
Founder @ Signal X
“Pretty fast and fun to use”
Nihir Kothari
Founder @ Sidekick Software
“It worked very well and provided accurate descriptions of the images we passed in. The API costs were very cheap.”
Sam Mayle
Researcher @ Mitsubishi Electric Research Lab
“It was faster than a lot of LLMs I've used, and it was easy to set up.”
Cedric Prentice
Software Engineer @ Wayfair
“Very cool, easy to use!”
Bill Simmons
Co-Founder @ Orbit.me
“Very interesting alternative to OpenAI”
Dave Gogi
Founder @ Signal X
“Pretty fast and fun to use”
Nihir Kothari
Founder @ Sidekick Software
“It worked very well and provided accurate descriptions of the images we passed in. The API costs were very cheap.”
Sam Mayle
Researcher @ Mitsubishi Electric Research Lab
“It was faster than a lot of LLMs I've used, and it was easy to set up.”
Cedric Prentice
Software Engineer @ Wayfair
“Very cool, easy to use!”
Bill Simmons
Co-Founder @ Orbit.me
“Very interesting alternative to OpenAI”
Dave Gogi
Founder @ Signal X
“Pretty fast and fun to use”
Nihir Kothari
Founder @ Sidekick Software
“It worked very well and provided accurate descriptions of the images we passed in. The API costs were very cheap.”
Sam Mayle
Researcher @ Mitsubishi Electric Research Lab
“Made the whole process very easy!”
Smruthi Ramesh
Lead Data Scientist @ Schneider Electric
“It was really fast!!”
Inder Singh
UDE
“Easy to use, great UI”
Yassine Fatimi
Founder @ ClauseGuard
“Pretty easy to use. No brainer. Easy drop in for OpenAI.”
Hansen Liang
Founder @ stealth
“Made the whole process very easy!”
Smruthi Ramesh
Lead Data Scientist @ Schneider Electric
“It was really fast!!”
Inder Singh
UDE
“Easy to use, great UI”
Yassine Fatimi
Founder @ ClauseGuard
“Pretty easy to use. No brainer. Easy drop in for OpenAI.”
Hansen Liang
Founder @ stealth
“Made the whole process very easy!”
Smruthi Ramesh
Lead Data Scientist @ Schneider Electric
“It was really fast!!”
Inder Singh
UDE
“Easy to use, great UI”
Yassine Fatimi
Founder @ ClauseGuard
“Pretty easy to use. No brainer. Easy drop in for OpenAI.”
Hansen Liang
Founder @ stealth
“Made the whole process very easy!”
Smruthi Ramesh
Lead Data Scientist @ Schneider Electric
“It was really fast!!”
Inder Singh
UDE
“Easy to use, great UI”
Yassine Fatimi
Founder @ ClauseGuard
“Pretty easy to use. No brainer. Easy drop in for OpenAI.”
Hansen Liang
Founder @ stealth
Want Subconscious inside your cloud?
For full control, run our inference system on your own GPUs. Use any model and harness, and increase the capacity of your GPU cluster.
Talk to an engineer- Models
- GLM, Nemotron, or any open weight model
- Compatibility
- Any coding harness
- Capacity
- Billions of tokens per day per node
- Pricing
- Per-GPU under management