Which LLM APIs and AI coding plans have a flat monthly price?
Synthetic, NanoGPT, Chutes, Ollama Cloud, Awan LLM and Standard Compute sell LLM API access for a fixed monthly price, and Z.ai, Kimi, MiniMax, OpenCode, Cline, Kilo and DevPass sell flat-price AI coding plans. None of them is unlimited: each caps requests, tokens, dollars of usage or time windows, and most cover open-weight models only.
Price, real limit and data region of 18 plans
| Option | Price a month | What you get | The catch | Works with | Data region |
|---|---|---|---|---|---|
| Standard Compute (us) | $19, $39, $89 or $249; Business from $499 | A fixed monthly compute budget of $20, $41, $95 or $269, spent at each model's per-token rate. Claude, GPT, Gemini and open models, routed per request or pinned. No 5-hour or weekly windows. | At 100% of the budget, new requests get HTTP 402 until renewal. Pay as you go (Economy and up) adds compute at the usage rate plus 15%. Pinning a premium model for every request uses the budget much faster. | OpenAI and Anthropic APIs. Claude Code, Codex CLI, OpenCode, Cline, Kilo Code, OpenClaw, Hermes Agent | EU endpoint by default, US endpoint optional. Providers worldwide unless you set a Europe-only or US-only rule |
| Synthetic | $30 a pack ($1 a day); packs stack | Per pack: 500 requests per 5 hours, $24 of API credit a week and 1 concurrent request per model, on 8 open-weight models (Kimi K3, GLM-5.3, DeepSeek V4.1 Flash, Qwen3.8 27B). | Three limits at once, and requests are weighted by model price (a Kimi K3 call counts as 1, a GLM-5.3 Flash call about 0.1). Open-weight models only. | OpenAI and Anthropic APIs. Claude Code, OpenCode, OpenClaw, Crush, Octofriend | Not published for inference; personal data processed in the US. API prompts not stored |
| NanoGPT | $12 | 60 million input tokens a week on 300+ open-weight models, plus 100 images a day. | Claude, GPT and Gemini are not in the subscription and bill per token. Inference runs at third-party providers, so retention varies by route. | OpenAI and Anthropic APIs. Claude Code, Codex CLI, OpenCode, Cline, Kilo Code, OpenClaw | Not published |
| Chutes | $10 (Plus) or $20 (Pro) | 2,000 or 5,000 requests a day on open-weight models running in TEEs, capped at 5x the price in pay-as-you-go value ($50 or $100 a month). | A 4-hour spend cap and the monthly value cap can stop you before the daily request count. Past the limits, requests bill at pay-as-you-go rates minus 6% or 10%. | OpenAI API (no Anthropic endpoint). Cline, Kilo Code, Roo Code, Cursor, Aider | Not published; API content not logged |
| Ollama Cloud | $20 (Pro) or $100 (Max) | $60 or $300 of usage credits a month at each model's token price, 3 or 10 concurrent requests, open-weight models only. | Metered since the switch to credits: Kimi K3 draws $3 per million input and $15 per million output tokens. Off-peak hours are cheaper. Credits don't roll over. | OpenAI and Anthropic APIs (a subset). Claude Code, Codex CLI, OpenCode, Cline, Hermes Agent, OpenClaw | Mainly US, may route to Europe. Prompts not logged |
| Featherless | $25 (Chat); Developer from $50 is per token | Chat plan: unmetered tokens on any open model in the catalogue, 32K context, 4 concurrency units (one 70B+ model at a time). | The flat plan is for chat you type yourself: no coding agents, no API traffic, no automation. Agents need the per-token Developer plan. | OpenAI API (Developer plan). Cline, Kilo Code, OpenCode, Pi | Not published. Prompts not stored |
| Awan LLM | $5, $10, $20 or $80 | Daily request quotas (Pro: 30,000 large-model requests a day; Max: no daily cap) and 20 to 200 requests a minute, tokens unmetered. | The model list is Llama 3 and 3.1 (8B and 70B) only, which is too weak for current coding agents. PayPal only. | OpenAI-style API; no coding agents documented | Not published; runs on its own servers |
| Z.ai GLM Coding Plan | $18, $80 or $168 (Lite, Pro, Max) | Credits: 2,000, 12,000 or 28,000 per 5 hours and 10,000, 60,000 or 140,000 a week, on GLM-5.3 and GLM-5.3 Flash. | Only inside supported coding tools, no general API calls. Weekdays 14:00 to 18:00 Singapore time burn credits at full rate; other hours at half. | OpenAI and Anthropic APIs. Claude Code, Codex, OpenCode, Cline, Kilo Code, Roo Code, Cursor | Not published for inference; personal data processed in Singapore |
| Kimi Code (Moonshot) | $19, $39, $99 or $199 (Plus to Ultra) | Kimi K3 and Kimi coding models with a rolling 5-hour window; amounts are not published, only multiples. | Interactive use only, no automation. A monthly membership cap freezes coding quota when reached. K3 with 1M context needs Pro or higher. | OpenAI and Anthropic APIs. Claude Code, OpenCode, Codex, Hermes Agent | Not published; the privacy policy allows training on user content |
| MiniMax M Plan | $22, $55 or $132 (Go, Explore, Build) | MiniMax M3.1 Flash Preview plus image and audio (video from Explore), at 1x, 3x or 7.5x usage; amounts not published. | 5-hour and weekly windows, peak-hour rate limits, and it is meant for individual interactive use, not production. | OpenAI and Anthropic APIs. Claude Code, Codex, Cursor, OpenCode, Hermes Agent, OpenClaw | Not published for inference; personal data stored in the US |
| OpenCode Go | $10 (Go) or $40 (Go Plus) | A dollar allowance per model each month (Go: $15 of GLM-5.3 or Kimi K3, $60 of MiniMax M3, $80 of DeepSeek V4.1 Flash), 20% of it per 5 hours and 50% per week. 31 models. | Frontier allowances are small ($15 of Claude Haiku 5.5 or GPT-6 Luna on Go). Single-member workspaces only, and traffic must look like a coding agent. | OpenAI and Anthropic APIs. OpenCode, Claude Code, Codex, Hermes Agent, Pi, Kilo Code CLI | Not published; open models keep 0 days of data, GPT, Grok and Claude 30 days |
| ClinePass | $9.99 | "2-5x the usage" of API rates on 12 open-weight models (GLM-5.3, Kimi K3, DeepSeek V4 Pro, MiniMax M3, Qwen3.8 Max). | 5-hour, weekly and monthly limits with no published amounts. Models get removed for capacity: four have been dropped so far. | OpenAI API. Built for Cline; the key also works in other agents | Not published |
| Kilo Pass | $19, $49 or $199 | The price as credits at providers' list prices on 500+ models, plus a bonus after the paid credits are used (up to $26.60, $68.60 or $278.60 a month in total). | Prepaid per-token credit, not an allowance: bonus credits expire monthly, the bonus streak resets if you cancel, and a 5% processing fee applies. | OpenAI API (Kilo Gateway). Kilo Code, OpenClaw, Hermes Agent | Hosted in the US |
| DevPass | $29, $79 or $179 | Usage worth 2x the price since Oct 15, 2026 ($58, $158 or $358; 3x before), metered at provider rates on 200+ models including Claude, GPT and Gemini. | Since Oct 15, 2026, a rolling 24-hour cap (8% to 10% of the allowance) and a weekly premium-model cap. Whitelisted coding tools only; API, automation or non-coding use can get the account banned. | OpenAI and Anthropic APIs. Claude Code, Codex, Cursor, Cline, OpenCode, OpenClaw, Hermes Agent | EU and US gateway processing; no EU-only option |
| Claude Pro / Max | $20, $100 or $200 | Claude in Claude Code and the Claude apps, per 5-hour session plus a weekly cap (Max: 5x or 20x Pro). Max also includes $100 or $200 of API credits a month. | Session amounts aren't published. The plan login only works in Claude Code and Anthropic's apps; other agents get only Max's API credits. Claude models only. | Plan: Claude Code and Anthropic apps. Max API credits: Anthropic API in any agent | Not published for consumer plans |
| ChatGPT Plus / Pro (Codex) | $20, or $100, $200 or $500 | Codex in the CLI, IDE, web and apps. Plus: an estimated 15 to 160 GPT-6.1 Sol messages per 5 hours; Pro: no 5-hour limit. | Weekly limits apply and aren't published. OpenAI models only. Other agents can use the plan only through the Sign in with ChatGPT preview for open-source apps. | Codex CLI, app and IDE extension; Responses API via Sign in with ChatGPT (preview) | Not published for Plus and Pro |
| GitHub Copilot | $10, $39 or $100 (Pro, Pro+, Max) | $15, $70 or $200 of AI credits a month (1 credit = $0.01) for chat, agent mode and CLI; code completions don't use credits. | Credits are metered per token, so agent sessions on frontier models drain them; extra usage bills at $0.01 a credit. Free, Pro and Pro+ interactions may be used for training unless you opt out. | Copilot in VS Code, JetBrains, CLI and GitHub; no base URL for other agents | Not published for individual plans |
| Cursor | $8, $30, $100 or $200 | A weekly pool of Grok usage (Premium 4x Plus, Super 15x, Ultra 50x), sold as a bundle with X Premium. | Claude, GPT and Gemini are not included and bill at API list price on top. Plus is Grok-only. The pool's size in dollars or tokens isn't published. | Cursor editor; Grok API billed separately | Not published for individual plans; US-only residency for Enterprise |
Prices in US dollars, checked on each vendor's own page on October 14, 2026. "Not published" means we could not find the value on any of the vendor's pages. Z.ai and Kimi prices come from data their pricing pages load, and Kimi publishes no quota amounts. Spot an error? Write to contact@standardcompute.com and we'll correct it.
Who each one fits
- Open-weight models in Claude Code or OpenCode, priced per request or per weekly token quota: Synthetic or NanoGPT.
- Light daily use of open models: Chutes, or Ollama Cloud if you already run Ollama locally.
- One vendor's model at a low price inside a supported coding tool: the GLM Coding Plan, Kimi Code or MiniMax M Plan, with windows and interactive-only terms.
- The cheapest pass for one agent: OpenCode Go in OpenCode, ClinePass in Cline.
- Any model at list price, billed per token with a bonus or a value cap: Kilo Pass or DevPass.
- Claude Code or Codex as your daily driver, within their session windows: Claude Max or ChatGPT Pro. Editor completions: GitHub Copilot.
- Chatting with open models yourself, not agents: Featherless Chat.
- Claude, GPT, Gemini and open models in any agent for one monthly price, with a hard budget cap and no windows: Standard Compute.
Head-to-heads between two plans
- ChatGPT Plus / Pro vs Claude Max, Cursor Premium / Ultra, DevPass, Featherless, GitHub Copilot Pro / Pro+ / Max
- Claude Max vs Cursor Premium / Ultra, DevPass, Featherless, GitHub Copilot Pro / Pro+ / Max
- Cursor Premium / Ultra vs DevPass, Featherless, GitHub Copilot Pro / Pro+ / Max
- DevPass vs Featherless, GitHub Copilot Pro / Pro+ / Max, OpenCode Go
- Featherless vs GitHub Copilot Pro / Pro+ / Max, OpenCode Go
- Cline Pass vs OpenCode Go, GLM Coding Plan (Z.ai), DevPass, Featherless
- GLM Coding Plan (Z.ai) vs OpenCode Go
The catch
Every plan in the table stops you somewhere: Synthetic at 500 requests per 5 hours, the GLM plan at its credit windows, DevPass at its daily cap, Claude Max at its session limit. Ours stops too. At 100% of the monthly compute budget, new requests get HTTP 402 until the plan renews, unless you upgrade or turn on Pay as you go (the usage rate plus 15%, Economy and up). Smart routing makes the budget last longer by sending routine steps to efficient models; pin a premium model for every request and it runs out much sooner.
Where Standard Compute fits
Standard Compute is the option here for closed frontier models (Claude, GPT, Gemini) and open models through one OpenAI- and Anthropic-compatible key, in any agent, with a monthly price that is the whole bill. Plans start at $19, have no 5-hour or weekly windows, and take an opt-in Europe-only or US-only rule per key.
FAQ
Is there a flat-rate LLM API I can plug into Claude Code, Cline or OpenCode?
Yes. Synthetic ($30), NanoGPT ($12), Ollama Cloud ($20) and Standard Compute (from $19) give you an API key with OpenAI- and Anthropic-compatible endpoints that work in all three. Chutes and Featherless's Developer plan are OpenAI-only. The Z.ai, Kimi and MiniMax plans also work in these agents but only for their own models, and Featherless's flat $25 plan excludes coding agents.
Which flat-price subscriptions include frontier models like Claude, GPT or Gemini?
Few. Synthetic, NanoGPT's subscription, Chutes, Ollama Cloud, Featherless and the Z.ai, Kimi and MiniMax plans are open-weight only. Claude Max covers Claude only and its login works only in Anthropic's apps; ChatGPT plans cover OpenAI only. Cursor bills Claude, GPT and Gemini at API list price on top. OpenCode Go includes small amounts of Claude Haiku 5.5 and GPT-6 Luna, DevPass and Kilo Pass meter them at provider rates, and Standard Compute includes Claude, GPT and Gemini in the monthly budget.
Is any flat-rate LLM API actually unlimited?
No plan for coding agents is. Featherless's $25 Chat plan has unmetered tokens but bans coding agents and API traffic, and Awan LLM's $80 Max plan has no daily request cap but serves only Llama 3 and 3.1. Every other plan caps requests, tokens, dollars of usage or time windows, and the table says which.
What happens when I hit the limit on a flat-rate plan?
It depends on the plan. Window plans (Claude, ChatGPT, Z.ai, Kimi, MiniMax, Synthetic, ClinePass) block you until the window resets. Chutes falls back to pay-as-you-go billing automatically, Copilot and Cursor bill extra usage if you allow it, and DevPass pauses unless you opt into overflow. Standard Compute returns HTTP 402 to new requests until the monthly renewal, and charges nothing extra unless you turn on Pay as you go.
Do these services train on my code or keep my prompts?
Synthetic, Chutes, Featherless and Ollama say they don't store or log API prompts. Kimi's privacy policy allows training on user content, and GitHub may train on Copilot Free, Pro and Pro+ interactions unless you opt out. On Standard Compute, keys without a rule follow each provider's own terms; a Europe-only or US-only rule requires zero data retention and no training at the provider.
Do flat-rate services serve quantized models?
Some say so. Ollama states its cloud models may use NVFP4 weights, and Synthetic lists an NVFP4 Nemotron; most others don't publish precision. On Standard Compute, Claude, GPT and Gemini are served by their makers or licensed clouds, and several open models run in FP8. If precision matters, pin a closed model.
What's the cheapest flat-rate option for a coding agent?
ClinePass ($9.99), OpenCode Go ($10), Chutes Plus ($10) and NanoGPT ($12) are the cheapest, all on open-weight models with tight limits. The cheapest plan that includes Claude, GPT and Gemini in its monthly price is Standard Compute Starter at $19 with a $20 compute budget.