Provider intelligence
Who's premium, who's cheap
Where each provider sits on price across the models it actually competes on (percentile 0 = always cheapest, 100 = always priciest). A provider consistently in the top quartile is selling something other than price — latency, compliance, brand.
Prices vendor-claimed · positions and percentiles are computed from them.
not yet measuredReliability, latency & uptime
We don't run latency/uptime/TTFT probes yet, so we have no measured reliability data — and we won't invent an SLA score from vendor marketing. When probes ship, a measured reliability axis appears here.
Build dependency: provider probe infrastructure (Phase 6)
126 providers
| Provider | Class | Models | Price position | Top-quartile | Open wts | Cache |
|---|---|---|---|---|---|---|
| Privatemode AI | relay / router | 11(8 comp.) | 92 | 88% | 100% | 100% |
| GreenPT | relay / router | 40(24 comp.) | 84 | 88% | 95% | 38% |
| Infomaniak | relay / router | 10(7 comp.) | 83 | 71% | 90% | 0% |
| above.dev | relay / router | 9(8 comp.) | 78 | 63% | 89% | 100% |
| Venice AI | — | 107(89 comp.) | 75 | 67% | 46% | 83% |
| Tinfoil | relay / router | 10(8 comp.) | 74 | 75% | 100% | 50% |
| Berget.AI | relay / router | 6(6 comp.) | 73 | 67% | 100% | 0% |
| Alibaba | first-party lab | 56(54 comp.) | 71 | 52% | 45% | 17% |
| STACKIT | relay / router | 8(5 comp.) | 69 | 60% | 100% | 0% |
| Scaleway | relay / router | 15(11 comp.) | 69 | 64% | 93% | 9% |
| Regolo AI | relay / router | 18(11 comp.) | 68 | 64% | 78% | 0% |
| Charm Hyper | relay / router | 34(23 comp.) | 67 | 43% | 74% | 100% |
| Inco | relay / router | 7(7 comp.) | 65 | 57% | 100% | 0% |
| Wafer | relay / router | 5(5 comp.) | 62 | 40% | 100% | 100% |
| evroc | relay / router | 16(12 comp.) | 61 | 58% | 94% | 0% |
| Mixlayer | relay / router | 5(5 comp.) | 60 | 20% | 100% | 0% |
| Cortecs | relay / router | 107(97 comp.) | 59 | 49% | 56% | 68% |
| Melious | relay / router | 15(15 comp.) | 56 | 40% | 93% | 100% |
| Groq | — | 16(9 comp.) | 56 | 33% | 75% | 33% |
| Clarifai | relay / router | 12(6 comp.) | 50 | 33% | 75% | 0% |
| Crusoe | relay / router | 11(11 comp.) | 50 | 27% | 100% | 100% |
| TensorX | relay / router | 32(32 comp.) | 49 | 34% | 91% | 100% |
| Cloudflare Workers AI | relay / router | 27(26 comp.) | 49 | 38% | 93% | 31% |
| AI-ROUTER | relay / router | 5(5 comp.) | 48 | 20% | 0% | 100% |
| Amazon Bedrock | hyperscaler | 165(128 comp.) | 47 | 34% | 31% | 60% |
| Fireworks AI | relay / router | 33(32 comp.) | 47 | 28% | 91% | 100% |
| Together AI | open-weights host | 39(35 comp.) | 46 | 23% | 87% | 51% |
| Jalapeno Cloud | relay / router | 17(17 comp.) | 43 | 12% | 88% | 0% |
| Nebius Token Factory | relay / router | 20(19 comp.) | 43 | 16% | 95% | 74% |
| Hugging Face | relay / router | 77(74 comp.) | 42 | 18% | 94% | 12% |
| Synthetic | relay / router | 10(10 comp.) | 42 | 30% | 100% | 100% |
| LLMTR | relay / router | 32(26 comp.) | 42 | 19% | 44% | 8% |
| Bothub | relay / router | 8(6 comp.) | 40 | 33% | 75% | 0% |
| Requesty | relay / router | 155(127 comp.) | 40 | 31% | 38% | 94% |
| OVHcloud AI Endpoints | relay / router | 15(12 comp.) | 39 | 17% | 100% | 0% |
| Volcengine Ark | relay / router | 16(16 comp.) | 38 | 31% | 25% | 100% |
| Baseten | relay / router | 23(22 comp.) | 37 | 9% | 96% | 64% |
| Friendli | relay / router | 7(7 comp.) | 37 | 14% | 100% | 86% |
| IteraCompute | relay / router | 11(9 comp.) | 37 | 22% | 100% | 100% |
| ai& | relay / router | 11(10 comp.) | 36 | 0% | 91% | 100% |
| OCI Generative AI | relay / router | 9(7 comp.) | 36 | 14% | 56% | 29% |
| Ollama Cloud | relay / router | 24(17 comp.) | 36 | 12% | 96% | 100% |
| UnoRouter | relay / router | 23(12 comp.) | 36 | 33% | 57% | 0% |
| DigitalOcean | relay / router | 97(39 comp.) | 34 | 18% | 48% | 74% |
| Vultr | relay / router | 10(9 comp.) | 34 | 33% | 100% | 0% |
| EmpirioLabs AI | relay / router | 62(56 comp.) | 34 | 20% | 60% | 100% |
| Zhipu AI | first-party lab | 16(14 comp.) | 33 | 7% | 88% | 86% |
| AKI.IO | relay / router | 8(7 comp.) | 32 | 14% | 100% | 43% |
| CoreWeave | relay / router | 27(23 comp.) | 32 | 9% | 96% | 100% |
| Neuralwatt | relay / router | 21(17 comp.) | 32 | 24% | 100% | 100% |
| Z.AI | relay / router | 17(15 comp.) | 31 | 7% | 88% | 87% |
| MiniMax (minimax.cn) | open-weights host | 7(7 comp.) | 31 | 14% | 100% | 86% |
| MiniMax (minimax.io) | first-party lab | 7(7 comp.) | 31 | 14% | 100% | 86% |
| SiliconFlow (China) | relay / router | 47(44 comp.) | 31 | 14% | 66% | 14% |
| ClinePass | relay / router | 15(15 comp.) | 31 | 7% | 80% | 100% |
| Alibaba (China) | relay / router | 90(77 comp.) | 31 | 17% | 47% | 22% |
| Chutes | relay / router | 14(13 comp.) | 30 | 15% | 93% | 100% |
| Ofox | relay / router | 144(143 comp.) | 29 | 15% | 30% | 96% |
| first-party lab | 39(26 comp.) | 29 | 15% | 5% | 65% | |
| NovitaAI | relay / router | 107(94 comp.) | 29 | 6% | 82% | 38% |
| Modelis | relay / router | 9(9 comp.) | 29 | 22% | 22% | 0% |
| Ambient | relay / router | 10(9 comp.) | 29 | 11% | 100% | 100% |
| 302.AI | relay / router | 117(106 comp.) | 28 | 17% | 25% | 3% |
| SiliconFlow | relay / router | 49(49 comp.) | 27 | 6% | 65% | 20% |
| Pioneer | relay / router | 112(94 comp.) | 27 | 15% | 51% | 100% |
| Vercel AI Gateway | — | 391(236 comp.) | 27 | 15% | 26% | 80% |
| NanoGPT | relay / router | 685(332 comp.) | 27 | 16% | 52% | 98% |
| Arcee | relay / router | 7(7 comp.) | 26 | 0% | 100% | 100% |
| CrossModel | relay / router | 60(60 comp.) | 26 | 12% | 33% | 98% |
| Meganova | relay / router | 19(18 comp.) | 26 | 11% | 79% | 0% |
| GMI Cloud | relay / router | 15(15 comp.) | 25 | 13% | 60% | 100% |
| ZenMux | relay / router | 128(107 comp.) | 25 | 12% | 36% | 80% |
| LLM Gateway | relay / router | 408(403 comp.) | 24 | 8% | 49% | 78% |
| HPC-AI | relay / router | 9(9 comp.) | 24 | 0% | 67% | 100% |
| TokenGo | relay / router | 13(13 comp.) | 23 | 0% | 100% | 92% |
| Abacus | relay / router | 108(104 comp.) | 23 | 6% | 28% | 45% |
| Eden AI | relay / router | 283(275 comp.) | 23 | 8% | 51% | 70% |
| OpenCode Go | relay / router | 36(34 comp.) | 23 | 3% | 53% | 100% |
| OpenCode Zen | relay / router | 107(73 comp.) | 22 | 8% | 36% | 99% |
| Impossibl | relay / router | 76(76 comp.) | 22 | 5% | 25% | 93% |
| AIHubMix | — | 82(65 comp.) | 22 | 8% | 38% | 100% |
| Databricks | relay / router | 30(24 comp.) | 22 | 4% | 13% | 92% |
| Deep Infra | open-weights host | 68(67 comp.) | 21 | 4% | 85% | 67% |
| IO.NET | relay / router | 17(14 comp.) | 21 | 7% | 88% | 100% |
| Vivgrid | — | 30(29 comp.) | 20 | 3% | 27% | 97% |
| Jiekou.AI | relay / router | 61(59 comp.) | 20 | 7% | 30% | 0% |
| Auriko | relay / router | 15(15 comp.) | 20 | 0% | 40% | 80% |
| Neon | relay / router | 46(45 comp.) | 19 | 7% | 26% | 82% |
| Merge Gateway | — | 187(184 comp.) | 19 | 5% | 44% | 72% |
| routing.run | relay / router | 15(12 comp.) | 19 | 17% | 67% | 0% |
| Perplexity Agent | — | 22(22 comp.) | 19 | 9% | 18% | 95% |
| Helicone | relay / router | 90(79 comp.) | 18 | 6% | 31% | 57% |
| OrcaRouter | relay / router | 117(108 comp.) | 18 | 3% | 32% | 81% |
| FrogBot | relay / router | 26(25 comp.) | 16 | 0% | 27% | 92% |
| FastRouter | relay / router | 47(36 comp.) | 16 | 0% | 30% | 25% |
| DInference | relay / router | 6(6 comp.) | 16 | 0% | 100% | 0% |
| DevPass (LLM Gateway) | relay / router | 194(190 comp.) | 16 | 4% | 34% | 74% |
| Vertex | hyperscaler | 52(47 comp.) | 16 | 6% | 21% | 89% |
| Azure Cognitive Services | — | 74(67 comp.) | 15 | 3% | 24% | 61% |
| Kilo Gateway | relay / router | 386(341 comp.) | 15 | 5% | 38% | 61% |
| Azure | hyperscaler | 89(81 comp.) | 15 | 5% | 24% | 63% |
| NEAR AI Cloud | relay / router | 37(30 comp.) | 14 | 3% | 24% | 87% |
| OpenRouter | — | 377(343 comp.) | 13 | 2% | 40% | 63% |
| SAP AI Core | — | 49(26 comp.) | 13 | 4% | 6% | 85% |
| Cloudflare AI Gateway | — | 47(46 comp.) | 13 | 4% | 6% | 89% |
| Mistral | first-party lab | 35(26 comp.) | 12 | 4% | 74% | 8% |
| Poe | relay / router | 137(72 comp.) | 12 | 7% | 11% | 72% |
| OpenAI | first-party lab | 48(41 comp.) | 12 | 2% | 0% | 78% |
| DaoXE | relay / router | 9(9 comp.) | 10 | 0% | 0% | 100% |
| Vancine | relay / router | 11(8 comp.) | 10 | 0% | 73% | 100% |
| Xpersona | relay / router | 13(12 comp.) | 10 | 8% | 0% | 100% |
| RunInfra | relay / router | 7(7 comp.) | 9 | 0% | 100% | 100% |
| FreeModel | — | 10(10 comp.) | 8 | 0% | 0% | 100% |
| Anthropic | first-party lab | 14(14 comp.) | 8 | 0% | 0% | 100% |
| Opper | relay / router | 40(40 comp.) | 8 | 3% | 13% | 80% |
| submodel | relay / router | 9(9 comp.) | 7 | 0% | 67% | 0% |
| GitHub Copilot | relay / router | 28(26 comp.) | 6 | 0% | 7% | 100% |
| CrofAI | relay / router | 24(19 comp.) | 5 | 0% | 83% | 100% |
| Vertex (Anthropic) | — | 14(14 comp.) | 2 | 0% | 0% | 100% |
| Xiaomi | first-party lab | 6(5 comp.) | 1 | 0% | 67% | 100% |
| QiHang | relay / router | 9(9 comp.) | 0 | 0% | 0% | 0% |
| xAI | first-party lab | 12(6 comp.) | 0 | 0% | 0% | 100% |
| StepFun (China) | first-party lab | 8(5 comp.) | 0 | 0% | 38% | 100% |
| Meta | first-party lab | 5(5 comp.) | 0 | 0% | 0% | 100% |
| StepFun (Global) | relay / router | 8(5 comp.) | 0 | 0% | 38% | 100% |
| Cohere | first-party lab | 14(5 comp.) | 0 | 0% | 100% | 0% |
Price position is the mean percentile of a provider's output $/M within each model it serves, over models with ≥ 2 providers. It is a positioning signal, not a quality judgement — and carries the same quant: unknown caveat as every price view.
Positioning quadrant
Price position against cache-pricing coverage (switch axes for open-weights share or catalog size). Bottom-left is cheap with rich cache pricing; the frontier marks providers no one dominates on both axes.
Positioned on published prices and catalog membership (vendor-claimed). There is deliberately no latency/reliability axis — we don't run probes yet, so we won't plot it.
Data: models.dev (MIT). Prices are vendor-published; verify before purchase. Methodology.