Everything we ship, change, and improve — new models, agent integrations, and performance work. Set your model to “standardcompute” and these land in your stack automatically.
Kilo Code is now supported. In VS Code, choose "OpenAI Compatible" as the provider, paste your base URL, API key, and the "standardcompute" model, and your editor agent runs on Standard Compute. A step-by-step guide is in the Dashboard under Agents.
Added official OpenCode support. Drop a Standard Compute provider into your opencode.json, set it as the default model, and you're running — no per-token billing and no separate key for each model. Setup guide available in the Dashboard.
Hermes Agent now connects to Standard Compute out of the box. Point it at our OpenAI-compatible base URL, set the model to "standardcompute," and the routing algorithm handles model selection from there.
This brings first-class agent integrations to four, alongside OpenClaw, OpenCode, and Kilo Code — all configured the same way, with no custom plugins.
A major upgrade to LLM routing. The system now scores each request on complexity, expected output length, and live provider load before selecting a model, and fails over across providers in under 200ms when one is degraded.
In our testing this improved output quality on demanding agentic tasks while lowering average latency on the Fast and Turbo tiers. No changes needed on your end — every integration benefits automatically.
The routing pool has been refreshed with the latest frontier models: GPT-5.1 and GPT-5.1 Codex, Claude Opus 4.7, Claude Sonnet 4.6, and Grok 4.2. You never pick a model — keep "standardcompute" set and the router chooses the best fit for each request. Existing integrations get the new models with no changes.
Standard Compute now supports OpenClaw, the open-source AI agent framework. A dedicated setup guide is available in the Dashboard with one-command installation for Mac, Linux, and Windows.
Set your model to "standardcompute" and our routing algorithm handles the rest — your agents get access to top-tier models from OpenAI, Anthropic, and xAI automatically.
Four new tiers: Starter ($9/mo), Standard ($39/mo), Fast ($99/mo), and Turbo ($399/mo). Every plan includes unlimited LLM compute. Higher tiers get faster execution and priority scheduling.
All plans include a 3-day free trial. Existing subscribers have been migrated automatically — no changes needed.
xAI's Grok 4 has been added to the routing pool alongside GPT-5.1 Codex and Claude Opus 4. Strong at tasks that benefit from real-time data and multi-agent reasoning. Selected automatically when it's the best fit.
We have published a fair use policy explaining how we keep unlimited compute sustainable — intelligent batching, LLM routing, prompt compaction, and adaptive throttling. Full details on the Fair Use page.
Anthropic's Claude Opus 4 is now in the routing pool. Excels at deep reasoning, multi-step planning, and long-running agentic tasks. Automatically selected for requests that benefit from extended context.
Improved priority scheduling for Fast and Turbo plans. Adaptive throttling is now more granular — graduated slowdowns instead of hard cutoffs, resulting in smoother performance during peak demand.
Standard Compute now officially supports Make and Zapier alongside n8n. All platforms work through the same OpenAI-compatible endpoints. Integration guides available on the Integrations page.
Major upgrade to our LLM routing. The system now considers request complexity, token budget, and current provider load to pick the optimal model. Cross-provider failover is automatic — if one provider has issues, your requests are seamlessly rerouted.
Deployed intelligent batching and smart prompt compaction across the platform. Batching groups requests to improve GPU utilization. Compaction strips redundant tokens before execution — 12–18% savings on average, often over 20% for templated automation workflows.
These systems, together with LLM routing, are the core of how we deliver unlimited compute at a flat price.
Standard Compute is live. Unlimited AI compute for a flat monthly price. OpenAI-compatible API with /v1/completions and /v1/responses endpoints.
We built this because automation teams were getting crushed by unpredictable per-token bills. Run as many AI-powered workflows as you need without worrying about cost. Welcome aboard.
The open beta period has concluded. All beta accounts have been migrated to the general availability platform with no downtime. API keys issued during beta remain valid — no action required.
Thank you to the 2,400+ teams who participated. Your feedback directly shaped the routing algorithm, throttling behavior, and pricing structure we shipped at launch.
OpenAI's GPT-5 is now available through Standard Compute. The routing algorithm selects it automatically for tasks where it outperforms alternatives — particularly multi-modal reasoning and structured data extraction.
Early preview of the Standard Compute dashboard. Manage API keys, view request logs, and monitor usage — all from a single interface. The dashboard is available to all beta users at standardcompute.com/dashboard.
This is a first version. Billing management, team controls, and usage analytics are coming in later releases.
Replaced the fixed rate-limit system with adaptive throttling. Instead of hard request caps, the system now gradually adjusts throughput based on real-time cluster load. During normal conditions, there are effectively no limits. During peak demand, lower-tier plans experience gentle slowdowns rather than hard rejections.
This eliminates the most common complaint from beta users — unexpected 429 errors during traffic spikes.
Published a step-by-step guide for connecting Standard Compute to n8n workflows. Point any OpenAI-compatible node at our endpoint, paste your API key, and you're running. No custom modules or plugins needed.
Added automatic cross-provider failover. If a request to one LLM provider fails or times out, the system retries against a different provider transparently. No changes needed on your end — failed requests are rerouted in under 200ms.
This was the number one reliability request during beta. In internal testing it reduced user-visible error rates by over 60%.
Anthropic's Claude 3.5 Sonnet is now in the routing pool. Particularly strong at code generation, technical writing, and instruction-following. The router favors it for automation prompts that require precise output formatting.
Standard Compute is now in open beta. Anyone can sign up, get an API key, and start sending requests. The beta is free — no credit card required. We're looking for feedback on latency, output quality, and the developer experience.
The core idea is simple: unlimited AI compute for a flat price. We handle model selection, infrastructure scaling, and cost optimization so you can focus on building.
The closed alpha has concluded after 12 weeks with 47 teams. Key outcomes: the routing algorithm now handles 14 model variants across 3 providers, average response latency is under 1.2 seconds, and the prompt compaction system is reducing token usage by 12–18% without measurable quality loss.
Every major issue surfaced during alpha has been addressed. We're preparing for open beta.
Standard Compute enters closed alpha with a small group of early partners. The initial release supports GPT-4o, Claude 3 Opus, and Gemini 1.5 Pro through a single OpenAI-compatible endpoint.
The goal for this phase is to validate the core routing and compaction systems under real workloads before opening access more broadly.