Auto routing
Auto serves the first model on your list within your price ceiling, then your default model.
AI inference.Up to 98% below OpenRouter.
Lower prices, one API, no subscriptions.
- Max discount
- 98%
- Models
- 339
- Model labs
- 63
- Routing overhead
- <100ms
| Model | OpenRouter | Fluence | Save |
|---|---|---|---|
| Qwen3.8 27BQwen | in$0.20out$2.50 | in$0.002out$0.022−98% | −98% |
| Grok 4.6xAI | in$2.00out$6.00 | in$0.612out$1.84−69% | −69% |
| Claude Sonnet 5.5Anthropic | in$2.00out$10.00 | in$0.88out$4.40−56% | −56% |
Pay less for the models you already run.
Compare our quoted prices with OpenRouter's current featured prices. Providers, features and retention can differ; supplier bills can adjust the final charge.
RetentionZDR
Zero data retention. Only providers that keep no copy of your prompts or replies, and never train on them. Send zdr: true on a request to require it. That pool is smaller than the full one, so these prices are usually higher than standard.
Read the docs| Model | OpenRouter | Fluence | Difference | Context |
|---|---|---|---|---|
| Qwen3.8 27Bqwen/qwen3.8-27b−98% | in$0.20out$2.50 | in$0.002out$0.022 | −98% | 262K |
| Grok 4.6x-ai/grok-4.6−69% | in$2.00out$6.00 | in$0.612out$1.84 | −69% | 200K |
| DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash−67% | in$0.30out$1.20 | in$0.096out$0.321 | −67% | 1M |
| Kimi K3moonshotai/kimi-k3−66% | in$0.239out$10.80 | in$0.08out$0.241 | −66% | 1M |
| Grok 4.7x-ai/grok-4.7−61% | in$2.00out$6.00in$4.00out$12.00 | in$0.774out$2.32in$1.55out$4.64 | −61% | 500K |
| GPT-5.6 Lunaopenai/gpt-5.6-luna−56% | in$0.20out$1.20in$0.40out$1.80 | in$0.088out$0.528in$0.22out$0.992 | −56% | 1M |
| GPT-6 Astraopenai/gpt-6-astra−56% | in$10.00out$50.00in$20.00out$75.00 | in$4.40out$22.00in$10.29out$38.57 | −56% | 1M |
| Claude Opus 5.5anthropic/claude-opus-5.5−56% | in$4.00out$20.00 | in$1.76out$8.80 | −56% | 1M |
| GPT-6.1 Solopenai/gpt-6.1-sol−56% | in$2.00out$10.00in$4.00out$15.00 | in$0.88out$4.40in$2.05out$7.70 | −56% | 1M |
| GLM 5.3 Flashz-ai/glm-5.3-flash−56% | in$0.15out$0.50 | in$0.066out$0.22 | −56% | 33K |
| Gemini 3.7 Flashgoogle/gemini-3.7-flash−56% | in$0.75out$3.75 | in$0.33out$1.65 | −56% | 33K |
| Claude Fable 5.1anthropic/claude-fable-5.1−56% | in$10.00out$50.00 | in$4.40out$22.00 | −56% | 33K |
| Llama 3.2 3B Instructmeta-llama/llama-3.2-3b-instruct−53% | in$0.051out$0.335 | in$0.024out$0.024 | −53% | 131K |
| Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b−50% | in$0.30out$2.40 | in$0.15out$0.75 | −50% | 262K |
| DeepSeek V4.1 Flash 1017deepseek/deepseek-v4.1-flash-1017−45% | in$0.15out$0.60 | in$0.081out$0.326 | −45% | 1M |
| Mistral Large 4mistralai/mistral-large-4-0−43% | in$1.36out$4.18 | in$0.763out$2.34 | −43% | 524K |
| Solar Pro 4upstage/solar-pro4−40% | in$0.15out$0.60 | in$0.09out$0.36 | −40% | 524K |
| Qwen2.5 Coder 32B Instructqwen/qwen-2.5-coder-32b-instruct−38% | in$0.66out$1.00 | in$0.404out$0.404 | −38% | 32K |
| Gemini 4 Argongoogle/gemini-4-argon−37% | in$2.00out$10.00 | in$1.24out$6.22 | −37% | 1M |
| Llama 3.1 Euryale 70B v2.2sao10k/l3.1-euryale-70b−37% | in$0.85out$0.85 | in$0.453out$0.528 | −37% | 20K |
| Aion-3.0aion-labs/aion-3.0−35% | in$3.00out$6.00 | in$1.94out$3.87 | −35% | 128K |
| Hermes 3 70B Instructnousresearch/hermes-3-llama-3.1-70b−35% | in$0.70out$0.70 | in$0.449out$0.449 | −35% | 66K |
| MiMo-V2.6-Pro-UltraSpeedxiaomi/mimo-v2.6-pro-ultraspeed−28% | in$4.35out$8.70 | in$3.11out$6.22 | −28% | 1M |
| Ember-1fireworks/ember-1−28% | in$3.00out$15.00 | in$2.15out$10.72 | −28% | 1M |
| Inklingthinkingmachines/inkling−28% | in$1.00out$4.05 | in$0.715out$2.90 | −28% | 524K |
| Seed 2.1 Turbobytedance-seed/seed-2-1-turbo−28% | in$0.50out$2.50 | in$0.357out$1.79 | −28% | 256K |
| Mercury 2inception/mercury-2−28% | in$0.25out$0.75 | in$0.179out$0.536 | −28% | 128K |
| Mistral Largemistralai/mistral-large−28% | in$2.00out$6.00 | in$1.43out$4.29 | −28% | 128K |
| Claude Haiku 5.5anthropic/claude-haiku-5.5−28% | in$0.10out$0.50 | in$0.071out$0.357 | −28% | 100K |
| Qwen3.8 Max Primeqwen/qwen3.8-max-prime−27% | in$4.00out$12.00 | in$2.91out$8.73 | −27% | 1M |
| Pareto 26.9unbiased/pareto-26.9−27% | in$2.50out$7.50 | in$1.82out$5.46 | −27% | 262K |
| Command A+cohere/command-a-plus−27% | in$0.30out$1.50 | in$0.218out$1.09 | −27% | 192K |
| LFM2.5-24B-A2Bliquid/lfm-2.5-24b-a2b−27% | in$0.05out$0.70 | in$0.036out$0.505 | −27% | 128K |
| Glimpseelorian/glimpse−24% | in$0.60out$3.00 | in$0.452out$2.26 | −24% | 262K |
| MiniMax M2.1minimax/minimax-m2.1−24% | in$0.30out$1.20 | in$0.226out$0.905 | −24% | 198K |
| Moondream 3.1moondream/moondream3.1-9b-a2b−21% | in$0.30out$2.50 | in$0.236out$1.97 | −21% | 32K |
| Fugu Ultra v2sakana/fugu-ultra-v2−20% | in$5.00out$30.00 | in$3.97out$23.84 | −20% | 272K |
| Ling 3.1 Flashinclusionai/ling-3.1-flash−20% | in$0.075out$0.225 | in$0.059out$0.178 | −20% | 262K |
| Hermes 4 405Bnousresearch/hermes-4-405b−20% | in$1.00out$3.00 | in$0.593out$2.37 | −20% | 128K |
| Llama 3.3 Euryale 70Bsao10k/l3.3-euryale-70b−16% | in$0.65out$0.75List price · current comparison unavailable | in$0.542out$0.542 | −16% | 20K |
| Hy3tencent/hy3−12% | in$0.083out$0.33 | in$0.073out$0.286 | −12% | 262K |
| WizardLM-2 8x22Bmicrosoft/wizardlm-2-8x22b−12% | in$0.62out$0.62List price · current comparison unavailable | in$0.542out$0.542 | −12% | 66K |
| Magnum v4 72Banthracite-org/magnum-v4-72b−11% | in$2.50out$5.00 | in$2.21out$3.29 | −11% | 16K |
| Muse Spark Openmeta/muse-spark-open−8% | in$1.25out$4.25 | in$1.15out$3.89 | −8% | 1M |
| GLM 5.3 FlashXz-ai/glm-5.3-flashx−7% | in$0.37out$1.25 | in$0.343out$1.16 | −7% | 1M |
| MiniMax M3.1 Flashminimax/minimax-m3.1-flash−6% | in$0.15out$0.60 | in$0.14out$0.561 | −6% | 1M |
| Step 5 Previewstepfun/step-5-preview−6% | in$1.00out$2.70 | in$0.935out$2.52 | −6% | 1M |
| Inkling Smallthinkingmachines/inkling-small−6% | in$0.45out$1.20 | in$0.281out$1.12 | −6% | 524K |
| Mercury 2.5inception/mercury-2.5−6% | in$0.04out$0.15 | in$0.037out$0.14 | −6% | 260K |
| Aion-3.0-Miniaion-labs/aion-3.0-mini−6% | in$0.70out$1.40 | in$0.654out$1.31 | −6% | 128K |
| Nemotron 3.5 Super VLnvidia/nemotron-3.5-super-vl−4% | in$0.15out$0.50 | in$0.143out$0.357 | −4% | 262K |
Security and routing without changing your stack.
Sensitive data redaction
Redact credentials and personal data before dispatch, or block the request.
Prototype
Tool-call protection
Block exfiltration, destructive actions, and instruction injection before delivery.
Scoped API keys
Set scopes, expiry, allowlists, rate limits, and budgets per key.
Zero data retention
Use verified zero-retention routes with caching and temporary uploads disabled.
Connect with your base URL and API key.
client = OpenAI(base_url="https://api.fluence.cloud/v1", api_key=os.environ["FLUENCE_API_KEY"])const client = new OpenAI({ baseURL: "https://api.fluence.cloud/v1", apiKey: process.env.FLUENCE_API_KEY });curl https://api.fluence.cloud/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $FLUENCE_API_KEY" \
-d '{"model": "deepseek/deepseek-v4-flash", "messages": [{"role": "user", "content": "Hi"}]}'- 1
Point your client at Fluence
Set the base URL to api.fluence.cloud/v1 in any OpenAI-compatible SDK, or in curl.
- 2
Paste one key
One key covers all 339 models, and you call each one by the id shown in the list.
- 3
Pay per token
You add funds from $5 and pay per token, with no monthly plan. Each request shows what it cost.
Import to your coding tool.
The tools you already use, connected to Fluence. Copy a setup prompt and let your agent handle the configuration.
- Hermes Agent
- Claude Code
- Kilo Code
- Cline
- pi
- omp
- Codex
- OpenCode
- OpenHands
- OpenClaw
- DeepSeek Harness
The discount comes from providers with spare capacity competing for your requests.
Spare capacity serves the same models
Idle GPUs host the open weights; providers with spare API capacity carry GPT, Claude and Gemini. All of it priced below OpenRouter.
Set routing per API key
A routing profile orders the models Auto may use, caps token prices, and picks cheaper or faster providers. Your privacy settings apply to every request.
OpenRouter's rate is the ceiling
With no better offer, you pay OpenRouter's public rate and nothing above it.
Fluence Cloud
Cloud VPS for AI from $2.54 a month.
Host your agents, APIs and workers on Fluence Cloud. It also rents dedicated VDS, block storage and on-demand GPUs.
Explore Fluence Cloud
from$2.54/mo2 vCPU · 2 GB RAM · 25 GB
Is it the same model?
Yes. Each row names the vendor's own model. Open-weight models run as the exact Hugging Face checkpoint; closed models go through the vendor's API, served by a provider with spare capacity.
Which models are on the list?
339 models with a public rate on OpenRouter, from OpenAI, Qwen, Google, Anthropic, Mistral, DeepSeek, Z.ai, Meta, MiniMax, Moonshot AI, xAI, AionLabs, ByteDance Seed, NVIDIA, Cohere, inclusionAI, Xiaomi, Amazon, Sakana, Inception, Nous, Sao10K, StepFun, TheDrummer, Upstage, IBM, Inference.net, Moondream, Morph, Nex AGI, Perceptron, Perplexity, Poolside, Relace, Tencent, Thinking Machines, Arcee AI, Baidu, Elorian, Fireworks, LiquidAI, Magnum v4 72B, Mancer, Meituan, Microsoft, Mistral Large, Mistral Large 2407, MythoMax 13B, Pareto, Pareto 26.10, Pareto 26.10 Preview, Pareto 26.9, PrismML, Qwen2.5 72B Instruct, Qwen2.5 Coder 32B Instruct, Reflection, Reka Edge, Reka Flash 3, ReMM SLERP 13B, TypeSafe, Venice, WizardLM-2 8x22B and Writer.
What changes in my code?
You change two values: the base URL (api.fluence.cloud/v1) and your API key. Requests and responses follow the OpenAI format, so your SDKs, agents and tools keep working.
How are prices set?
Provider offers have their own rates. Your routing profile ranks them by price or by measured speed, within your privacy settings. You pay the provider's verified bill plus the model markup, with price matching when it covers supplier cost. For a delayed supplier bill, quoted rates apply first; after three unsuccessful checks, the charge moves to the captured full list price for actual usage. That charge stays provisional while Fluence checks for the bill for 30 days. A later bill corrects the charge with a debit or credit in your balance history. Some models charge more for long prompts; the price table marks their thresholds. A valid actual bill can exceed the quote or reference rate.
What do you keep from my requests?
We store the model, the token counts, the cost and the time, and none of the prompt or completion text. Pick a zero data retention route when your policy requires it.
When do I get access?
We send invites in waves, in signup order, and you get one email when yours is ready.
