InferenceDirect passes through provider inference pricing exactly as billed — no markup on the per-token rate. Prepaid balance, no monthly fee, no surprise invoices. Every model below is served through one OpenAI-compatible endpoint at api.inferencedirect.com.
The Claude family — frontier reasoning, coding, and long-context models.
| Model | Context | Input | Output |
|---|---|---|---|
| Claude Opus 4 | 200K | $15.00 | $75.00 |
| Claude Fable 5 | 200K | $11.00 | $55.00 |
| anthropic/claude-fable-5 | 1000K | $10.00 | $50.00 |
| Claude Opus 4.5 | 200K | $5.50 | $27.50 |
| Claude Opus 4.6 | 200K | $5.50 | $27.50 |
| Claude Opus 4.7 | 200K | $5.50 | $27.50 |
| Claude Opus 4.8 | 200K | $5.50 | $27.50 |
| Claude Opus 4.5 | 200K | $5.00 | $25.00 |
| anthropic/claude-opus-4-7 | 1000K | $5.00 | $25.00 |
| anthropic/claude-opus-4-8 | 1000K | $5.00 | $25.00 |
| Claude Sonnet 4.5 | 200K | $3.30 | $16.50 |
| Claude Sonnet 4.6 | 200K | $3.30 | $16.50 |
| Claude 3 Sonnet | 200K | $3.00 | $15.00 |
| Claude Sonnet 4 | 200K | $3.00 | $15.00 |
| Claude Sonnet 4 | 200K | $3.00 | $15.00 |
| Claude Sonnet 4 (OpenRouter) | 200K | $3.00 | $15.00 |
| Claude Sonnet 4.5 | 200K | $3.00 | $15.00 |
| anthropic/claude-sonnet-4-6 | 1000K | $3.00 | $15.00 |
| Claude Sonnet 5 | 200K | $2.20 | $11.00 |
| anthropic/claude-sonnet-5 | 1000K | $2.00 | $10.00 |
| Claude Haiku 4.5 | 200K | $1.10 | $5.50 |
| Claude Haiku 4.5 | 200K | $1.00 | $5.00 |
| anthropic/claude-haiku-4-5 | 200K | $1.00 | $5.00 |
| Claude 3.5 Haiku | 200K | $0.80 | $4.00 |
| Claude 3 Haiku | 200K | $0.25 | $1.25 |
| Claude 3 Haiku | 200K | $0.25 | $1.25 |
Amazon's own Nova and Titan models, served natively on Bedrock.
| Model | Context | Input | Output |
|---|---|---|---|
| Nova Pro | 300K | $1.05 | $4.20 |
| Nova 2.0 Lite | 300K | $0.43 | $3.60 |
| Nova Lite | 300K | $0.078 | $0.31 |
| Nova Micro | 128K | $0.046 | $0.18 |
Open-weight reasoning and general models with standout cost-efficiency — strong at code, math, and long-form reasoning.
| Model | Context | Input | Output |
|---|---|---|---|
| deepseek-ai/DeepSeek-V4-Pro | 1049K | $1.30 | $2.60 |
| deepseek-ai/DeepSeek-R1-0528 | 164K | $0.50 | $2.15 |
| deepseek-ai/DeepSeek-V3 | 164K | $0.32 | $0.89 |
| DeepSeek Chat (OpenRouter) | 64K | $0.28 | $0.88 |
| deepseek-ai/DeepSeek-V3.1-Terminus | 164K | $0.27 | $0.95 |
| deepseek-ai/DeepSeek-V3.2 | 164K | $0.26 | $0.38 |
| deepseek-ai/DeepSeek-V3.1 | 164K | $0.25 | $0.95 |
| deepseek-ai/DeepSeek-V3-0324 | 164K | $0.24 | $0.90 |
| deepseek-ai/DeepSeek-V4-Flash | 1049K | $0.09 | $0.18 |
The Qwen family — dense and MoE variants covering chat, coding, and vision.
| Model | Context | Input | Output |
|---|---|---|---|
| Qwen/Qwen3.7-Max | 256K | $2.50 | $7.50 |
| Qwen/Qwen3-Max | 256K | $1.20 | $6.00 |
| Qwen/Qwen3-Max-Thinking | 256K | $1.20 | $6.00 |
| Qwen/Qwen3.5-397B-A17B | 262K | $0.45 | $3.00 |
| Qwen/Qwen2.5-72B-Instruct | 33K | $0.36 | $0.40 |
| Qwen/Qwen3.6-27B | 262K | $0.32 | $3.20 |
| Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo | 262K | $0.30 | $1.00 |
| Qwen/Qwen3.5-122B-A10B | 262K | $0.29 | $2.40 |
| Qwen/Qwen3.5-27B | 262K | $0.26 | $2.60 |
| Qwen/Qwen3-235B-A22B-Thinking-2507 | 262K | $0.23 | $2.30 |
| Qwen/Qwen3-VL-235B-A22B-Instruct | 262K | $0.20 | $0.88 |
| Qwen/Qwen3-VL-30B-A3B-Instruct | 262K | $0.15 | $0.60 |
| Qwen/Qwen3.6-35B-A3B | 262K | $0.15 | $0.95 |
| Qwen/Qwen3.5-35B-A3B | 262K | $0.14 | $1.00 |
| Qwen/Qwen3-14B | 41K | $0.12 | $0.24 |
| Qwen/Qwen3-30B-A3B | 41K | $0.12 | $0.50 |
| Qwen/Qwen3.5-9B | 262K | $0.10 | $0.15 |
| Qwen/Qwen3-235B-A22B-Instruct-2507 | 262K | $0.09 | $0.55 |
| Qwen/Qwen3-Next-80B-A3B-Instruct | 262K | $0.09 | $1.10 |
| Qwen/Qwen3-32B | 41K | $0.08 | $0.28 |
| Qwen/Qwen3-Embedding-4B | 33K | $0.02 | $0.00 |
| Qwen/Qwen3-Embedding-0.6B | 33K | $0.01 | $0.00 |
| Qwen/Qwen3-Embedding-8B | 33K | $0.01 | $0.00 |
Meta's open-weight Llama line — dependable general-purpose instruct models with a deep tooling ecosystem.
| Model | Context | Input | Output |
|---|---|---|---|
| NousResearch/Hermes-3-Llama-3.1-405B | 131K | $1.00 | $1.00 |
| NousResearch/Hermes-3-Llama-3.1-70B | 131K | $0.70 | $0.70 |
| meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo | 131K | $0.40 | $0.40 |
| meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 | 1049K | $0.20 | $0.80 |
| Llama 3.2 3B | 128K | $0.19 | $0.19 |
| meta-llama/Llama-Guard-4-12B | 164K | $0.18 | $0.18 |
| Llama 3.2 1B | 128K | $0.13 | $0.13 |
| Llama 3.3 70B (OpenRouter) | 128K | $0.12 | $0.30 |
| meta-llama/Llama-3.3-70B-Instruct-Turbo | 131K | $0.10 | $0.32 |
| meta-llama/Llama-4-Scout-17B-16E-Instruct | 328K | $0.10 | $0.30 |
| meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo | 131K | $0.02 | $0.03 |
OpenAI's gpt-oss open-weight releases — Apache-licensed reasoning models in efficient MoE sizes.
| Model | Context | Input | Output |
|---|---|---|---|
| GPT-4o (OpenRouter) | 128K | $2.50 | $10.00 |
| openai/gpt-oss-120b-Turbo | 131K | $0.15 | $0.60 |
| openai/gpt-oss-120b | 131K | $0.037 | $0.17 |
| openai/gpt-oss-20b | 131K | $0.03 | $0.14 |
Efficient European open-weight models — small, fast instruct models that punch above their size.
| Model | Context | Input | Output |
|---|---|---|---|
| Pixtral Large 25.02 | 128K | $2.00 | $6.00 |
| mistralai/Mistral-Small-3.2-24B-Instruct-2506 | 128K | $0.075 | $0.20 |
| mistralai/Mistral-Small-24B-Instruct-2501 | 33K | $0.05 | $0.08 |
| mistralai/Mistral-Nemo-Instruct-2407 | 131K | $0.019 | $0.03 |
Nemotron models — open reasoning lines tuned for enterprise serving.
| Model | Context | Input | Output |
|---|---|---|---|
| nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B | 262K | $0.50 | $2.20 |
| nvidia/Nemotron-Content-Safety-3.5 | 131K | $0.20 | $0.20 |
| nvidia/NVIDIA-Nemotron-3-Super-120B-A12B | 262K | $0.085 | $0.40 |
| nvidia/Nemotron-3-Nano-30B-A3B | 262K | $0.05 | $0.20 |
| nvidia/llama-nemotron-embed-vl-1b-v2 | 10K | $0.01 | $0.00 |
MoE generalist models balancing strong agentic performance with aggressive pricing.
| Model | Context | Input | Output |
|---|---|---|---|
| MiniMaxAI/MiniMax-M2.7-Turbo | 197K | $0.38 | $1.70 |
| MiniMaxAI/MiniMax-M3 | 524K | $0.30 | $1.20 |
| MiniMaxAI/MiniMax-M2.7 | 197K | $0.25 | $1.00 |
The GLM family — frontier-level coding and agentic performance with fast, inexpensive Flash tiers.
| Model | Context | Input | Output |
|---|---|---|---|
| zai-org/GLM-5.1 | 203K | $1.05 | $3.50 |
| zai-org/GLM-5.2 | 1049K | $0.93 | $3.00 |
| zai-org/GLM-5 | 203K | $0.60 | $2.08 |
| zai-org/GLM-4.6 | 203K | $0.50 | $2.00 |
| zai-org/GLM-4.7 | 203K | $0.40 | $1.75 |
| zai-org/GLM-4.7-Flash | 203K | $0.06 | $0.40 |
Gemma open models — compact, efficient instruct models.
| Model | Context | Input | Output |
|---|---|---|---|
| google/gemini-3.1-pro | 1000K | $2.00 | $12.00 |
| google/gemini-3.5-flash | 1000K | $1.50 | $9.00 |
| Gemini 2.5 Pro (OpenRouter) | 1000K | $1.25 | $10.00 |
| google/gemini-2.5-flash | 1000K | $0.30 | $2.50 |
| google/gemini-3.1-flash-lite | 1000K | $0.25 | $1.50 |
| google/gemma-4-31B-it | 262K | $0.13 | $0.38 |
| google/gemma-4-31B-it-turbo | 262K | $0.12 | $0.37 |
| google/gemma-3-27b-it | 131K | $0.08 | $0.16 |
| google/gemini-1.5-flash | 1000K | $0.075 | $0.30 |
| google/gemma-4-26B-A4B-it | 262K | $0.07 | $0.34 |
| google/gemma-3-12b-it | 131K | $0.05 | $0.15 |
| google/gemma-3-4b-it | 131K | $0.05 | $0.10 |
| google/gemini-1.5-flash-8b | 1000K | $0.037 | $0.15 |
| google/gemma-4-E4B-it | 131K | $0.02 | $0.10 |
| google/embeddinggemma-300m | 2K | $0.002 | $0.00 |
Available through the unified endpoint.
| Model | Context | Input | Output |
|---|---|---|---|
| GPT-4o | 128K | $2.50 | $10.00 |
| GPT-4o (Azure) | 128K | $2.50 | $10.00 |
| GPT-4.1 | 1000K | $2.00 | $8.00 |
| GPT-4.1 (Azure) | 1000K | $2.00 | $8.00 |
| o3 | 200K | $2.00 | $8.00 |
| o3 (Azure) | 200K | $2.00 | $8.00 |
| Gemini 2.5 Pro | 1000K | $1.25 | $10.00 |
| o4-mini | 200K | $1.10 | $4.40 |
| o4-mini (Azure) | 200K | $1.10 | $4.40 |
| XiaomiMiMo/MiMo-V2.5-Pro | 1049K | $1.00 | $3.00 |
| thinkingmachines/Inkling | 131K | $1.00 | $4.05 |
| Sao10K/L3.1-70B-Euryale-v2.2 | 131K | $0.85 | $0.85 |
| moonshotai/Kimi-K2.6 | 262K | $0.75 | $3.50 |
| moonshotai/Kimi-K2.7-Code | 262K | $0.74 | $3.50 |
| ByteDance/Seed-2.0-code | 256K | $0.50 | $3.00 |
| ByteDance/Seed-2.0-pro | 256K | $0.50 | $3.00 |
| moonshotai/Kimi-K2.5 | 262K | $0.45 | $2.25 |
| GPT-4.1 mini | 1000K | $0.40 | $1.60 |
| Gryphe/MythoMax-L2-13b | 4K | $0.40 | $0.40 |
| XiaomiMiMo/MiMo-V2.5 | 262K | $0.40 | $2.00 |
| Gemini 2.5 Flash | 1000K | $0.30 | $2.50 |
| ByteDance/Seed-1.8 | 256K | $0.25 | $2.00 |
| stepfun-ai/Step-3.7-Flash | 262K | $0.20 | $1.15 |
| GPT-4o mini | 128K | $0.15 | $0.60 |
| GPT-4o mini (Azure) | 128K | $0.15 | $0.60 |
| tencent/Hy3 | 262K | $0.14 | $0.58 |
| ByteDance/Seed-2.0-mini | 256K | $0.10 | $0.40 |
| GPT-4.1 nano | 1000K | $0.10 | $0.40 |
| Gemini 2.0 Flash | 1000K | $0.10 | $0.40 |
| Gemini 2.5 Flash-Lite | 1000K | $0.10 | $0.40 |
| microsoft/phi-4 | 16K | $0.07 | $0.14 |
| Sao10K/L3-8B-Lunaris-v1-Turbo | 8K | $0.04 | $0.05 |
| BAAI/bge-en-icl | 8K | $0.01 | $0.00 |
| BAAI/bge-large-en-v1.5 | 1K | $0.01 | $0.00 |
| BAAI/bge-m3 | 8K | $0.01 | $0.00 |
| BAAI/bge-m3-multi | 8K | $0.01 | $0.00 |
| intfloat/e5-large-v2 | 1K | $0.01 | $0.00 |
| intfloat/multilingual-e5-large | 1K | $0.01 | $0.00 |
| intfloat/multilingual-e5-large-instruct | 1K | $0.01 | $0.00 |
| thenlper/gte-large | 1K | $0.01 | $0.00 |
| BAAI/bge-base-en-v1.5 | 1K | $0.005 | $0.00 |
| intfloat/e5-base-v2 | 1K | $0.005 | $0.00 |
| sentence-transformers/all-MiniLM-L12-v2 | 1K | $0.005 | $0.00 |
| sentence-transformers/all-MiniLM-L6-v2 | 1K | $0.005 | $0.00 |
| sentence-transformers/all-mpnet-base-v2 | 1K | $0.005 | $0.00 |
| sentence-transformers/clip-ViT-B-32 | — | $0.005 | $0.00 |
| sentence-transformers/clip-ViT-B-32-multilingual-v1 | 1K | $0.005 | $0.00 |
| sentence-transformers/multi-qa-mpnet-base-dot-v1 | 1K | $0.005 | $0.00 |
| sentence-transformers/paraphrase-MiniLM-L6-v2 | 1K | $0.005 | $0.00 |
| shibing624/text2vec-base-chinese | 1K | $0.005 | $0.00 |
| thenlper/gte-base | 1K | $0.005 | $0.00 |
● Prices are per 1 million tokens, served via AWS Bedrock, at exact provider cost — no markup. Embedding, image, and audio models bill per provider unit and are not listed here.
Free to start. Metered usage against a prepaid balance, personal budgets, and the full dashboard.
Prepaid balance plus governance: department and team budgets, warn/block enforcement, per-model policy, service keys, and chargeback exports.
Custom governance: strict reserve-then-settle enforcement, delegated administration, fiscal-period alignment. Directory sync is a future conversation, not a shipped feature.
provider usage at exact provider cost = metered cost, drawn from your account's prepaid balance. No per-token markup, ever.
Budgets cap the metered cost per member, team, and department. The ledger records both customer price and provider cost, so finance can reconcile to the cent.
Our revenue comes from a flat 5.5% platform fee ($0.80 minimum) charged once when you add funds by card — 5% for crypto, no minimum. The fee is added on top of the amount you choose and shown as its own line item at checkout, so a $10 top-up credits your balance the full $10 and charges your card $10.80 (the $0.80 minimum applies, since 5.5% of $10 is only $0.55). $10 minimum top-up, $25,000 maximum per transaction.
Cryptographic request signing (Ed25519, per tenant / user / node) plus per-request access policy — required claims, time-of-day windows, JSON schema validation, and mandatory signing. Prepaid: $5 per 1,000,000 calls, billed only while enabled.
Usage analytics and chargeback reporting with per-user and per-team drill-down and CSV export. Usage-based pricing drawn from your prepaid balance; storage scales from D1 to R2 for high volume.
Add-ons are independent — fund either, both, or neither on the same prepaid balance.
Top up by card (Stripe) or crypto (USDC via Coinbase) and set balance auto-top-up from a pre-authorised allowance — pay-as-you-go from a prepaid balance, with no subscription. Administrators can also enter manual credit.