Pricing
Start free. Pay only for what you use.
10+ open-source models are $0 input/output, capped only by the free-plan daily request limit. Premium providers (OpenAI, Anthropic, Google) are billed per token against your balance.
Free
Evaluate and prototype with open-source models.
- Ongoing NVIDIA NIM open-source access after a $5 activation (the $5 becomes spendable balance)
- 10+ open-source models on NVIDIA NIM
- Nemotron, GLM 5.3, gpt-oss-20b, and more
- OpenAI-compatible and Anthropic-compatible endpoints
- Automatic failover across providers
Activation
Cheapest path to unlock ongoing free use and any paid provider.
- Ongoing $0 input/output access to 10+ NVIDIA NIM open models (daily request limits apply)
- $5 becomes spendable balance — not a fee, not a subscription
- Plus $5/month of premium-provider usage included from then on (OpenAI, Anthropic, Google)
- OpenAI, Anthropic, Google, Replicate
- One-time charge, no recurring billing
Starter
Higher daily limits for free NVIDIA NIM models.
- 2,000 chat requests/day on 10+ free NVIDIA NIM models
- No premium-provider credit included in the $9 monthly subscription
- Premium calls use prepaid credit added separately
- OpenAI, Anthropic, Google, Replicate
- Requests stop at the daily limit; metered overages are not currently available
Pro· Most popular
For developers shipping production AI features.
- $20/month of premium-provider usage included
- Higher daily request limits (5,000 chat / 10,000 text / 500 image-generate)
- All 100+ models across every provider
- Image generation (Flux, DALL·E, Imagen)
- Video generation (Kling, Runway, Veo)
Capacity
Reserved concurrency on hardware we actually own.
- Reserved concurrent streams on our 4×3090 rig (96GB VRAM)
- Priority over all shared traffic — predictable latency under load
- 70B-class open models at the quality sweet spot
- Everything in Pro included
- The honest fit: low-volume, high-value workloads — not cheap bulk inference
Enterprise
Custom included usage, dedicated support, and compliance.
- Custom monthly included usage and request limits
- Custom model hosting
- Dedicated account team
- Priority support SLA
- Custom DPA and compliance
Token prices
Prices shown are the providers' own published rates per 1M tokens (input / output), checked September 2026; see each provider's website for the current rate. InferAll bills these at the provider's rate plus 30%, or 15% below it on Anthropic models, with no per-request minimum; the rate you pay for every model is listed on inferall.ai/models.
See inferall.ai/live for the full model list and pricing.
Common questions
Is the free plan actually free?
Yes. The 10+ open-source models on NVIDIA NIM are $0 input and $0 output, capped only by the free-plan daily request rate (50 chat / 100 text / 20 image-analyze / 10 image-generate / 5 video-generate per day). New accounts get 25 free NIM calls before any payment; after those, the $5 starter pack unlocks ongoing NIM access. That $5 is not a fee, it becomes spendable balance on your account and covers premium-model usage. Premium providers (OpenAI, Anthropic, Google) are billed per token, drawn from your balance or your plan's monthly included usage.
How does the monthly included usage work?
Monthly included premium-provider usage is $5/month on activated Free and $20/month on Pro. Starter has no monthly included premium credit; premium calls draw from prepaid credit added separately. Capacity sells reserved GPU concurrency rather than a token allowance. Calls to premium providers (OpenAI, Anthropic, Google) draw against your included usage first, then against any prepaid balance you've added via the $5 or $100 packs. Calls to the 10+ NVIDIA NIM open models cost $0 and don't burn included usage. Included usage resets monthly.
What does Starter include, and what happens at the daily limit?
The $9/month Starter subscription raises the chat limit to 2,000 requests/day. Free NVIDIA NIM models stay $0 input/output within that limit. Premium calls require prepaid credit added separately; the subscription includes no premium credit. Daily limits apply to both free and premium requests. Once a limit is reached, further requests for that operation stop until it resets. Metered overages are not currently available.
What's the difference between Free, the $5 activation pack, and Pro?
Free lets you try the 10+ NVIDIA NIM open-source models with a short no-card trial. The $5 activation pack is a one-time charge that unlocks ongoing free-plan use and becomes spendable balance for premium providers (OpenAI, Anthropic, Google) billed per token. Pro ($29/mo) gives you higher daily request limits and $20/month of premium-provider usage included — the right fit if you call paid providers regularly.
Claude Opus 4 is $15/$75 per million tokens — what's the free alternative?
For Opus-class work (long-context reasoning, agentic tasks, complex analysis), the closest free-NIM alternative is nvidia/nemotron-3-ultra-550b-a55b — a 550B-parameter model (55B active) with comparable quality on most benchmarks, billed at $0 input / $0 output. A typical Opus conversation (10K input + 4K output tokens) costs ~$0.45 on Anthropic's published Opus 4.7 rate; the same conversation through Nemotron 3 Ultra is $0. The $5 starter pack is most useful as a way to keep premium Opus available on-demand for the few tasks where it matters, with NIM handling the bulk. For smaller or latency-sensitive tasks, fetch GET https://api.inferall.ai/ai/v1/models and compare the current $0 IDs; this FAQ names Nemotron because the production hot-route probe currently returns HTTP 200 for it. NVIDIA retires NIM models on its own schedule, so treat any roster on this page as a starting point rather than a permanent promise.