# InferKey Review Prepaid API gateway giving developers one key for Claude and GPT models at up to 81% below official prices. ## Editor's Score: 62/100 Capability scores 16 because InferKey does one thing (relay tokens at a discount) and does it across a wide model list, but it has no team management, no failover, no audit logging, and no routing logic that any of the named competitors offer in their broader tooling. Ease of use scores 20 because the migration path is genuinely minimal: one base URL change and one key swap, with dual SDK compatibility confirmed on the pricing page. Value scores 18 because the discounts are real and substantial (49% on most Claude models, up to 81% on some GPT models) and the no-expiry prepaid model suits irregular usage, but the complete absence of a free trial or starter credit is a concrete barrier that costs points. Delivery scores 8 because no external user reviews, press coverage, or community discussion about InferKey was found in web search, the model names listed on the pricing page (gpt-5.4, gpt-6-luna, claude-fable-5) do not match any publicly announced model names, and the mechanism by which a small independent gateway achieves 49% to 81% discounts on official frontier models is not explained, all of which are material trust concerns a buyer cannot resolve before purchasing. ## Pros - Savings of 23% to 81% below official per-token prices, with the rate locked at purchase time so upstream price hikes do not affect pre-bought credits. - Dual-endpoint compatibility (OpenAI and Anthropic SDKs both work) means zero code changes beyond swapping the base URL and API key. - No subscription and no credit expiry: credits sit on the account until used, which suits irregular or burst usage patterns without wasted spend. ## Cons - No team or multi-user workspace: there is no way to assign per-key budgets, set spend limits per contributor, or separate usage by project, which makes it impractical for any team larger than one. - No provider failover or redundancy: if upstream model capacity is constrained, requests simply fail until capacity frees up, with no automatic rerouting to an alternative provider. - No free tier or trial credits: a developer cannot test the gateway's latency or verify output fidelity before committing real money, unlike some competing gateways that offer a small free credit on signup. ## Pricing Paid. - claude-fable-5: $5.13 input / $25.65 output per 1M tokens - claude-haiku-4-5: $0.51 input / $2.55 output per 1M tokens - claude-opus-4 (variants 6, 7, 8) and claude-opus-5: $2.55 input / $12.84 output per 1M tokens - claude-sonnet-5: $1.53 input / $7.68 output per 1M tokens - claude-sonnet-4-6: $1.53 input / $7.68 output per 1M tokens - gpt-5.4: $0.51 input / $3.00 output per 1M tokens - gpt-5.4-mini: $0.15 input / $0.90 output per 1M tokens - gpt-5.5: $0.99 input / $6.00 output per 1M tokens - gpt-6-luna: $0.03 input / $0.09 output per 1M tokens - gpt-5.6-luna: $0.03 input / $0.24 output per 1M tokens - gpt-5.6-terra: $0.51 input / $3.00 output per 1M tokens - gpt-5.6-sol: $0.99 input / $6.00 output per 1M tokens - gpt-image-2 (image generation): $0.0177 to $0.0354 per image Pricing reflects what we saw at the time of review (2026-11). Confirm current pricing on the official site. Category: AI Coding & Dev Tools (https://aitoolseekers.com/category/ai-coding-dev-tools) Official site: https://inferkey.dev/ Last verified by a human: 2026-11-18 Reviewed by a human at AI Tool Seekers, a hand-reviewed AI tools directory. Every listed tool is verified by a person before publishing. Scores are editorial opinions from a fixed rubric (capability, ease of use, value, delivery on promise) and cannot be bought; there is no pay-to-rank. Directory: https://aitoolseekers.com This review: https://aitoolseekers.com/tools/inferkey Methodology: https://aitoolseekers.com/how-verification-works ## Verdict InferKey is the right pick for solo developers and indie hackers who want discounted access to frontier models without committing to a monthly subscription. The tradeoff is that it is a thin relay with no team management, no audit logging, and no routing logic, so it loses ground to more full-featured gateways the moment a second person joins the project. Teams or anyone who needs spend controls per user, SSO, or provider failover should look elsewhere. Best for: Solo developers cutting LLM API costs with no subscription Not for: Teams needing per-seat spend controls, audit logs, or provider failover ## Overview InferKey is a prepaid LLM API gateway that puts Claude and GPT models behind a single API key. The core pitch is simple: buy credits, swap your base URL and API key in existing SDK code, and call the same frontier models at a discount. The site claims savings of 23% to 81% below official prices depending on the model, with rates locked at the time of purchase so a price increase at Anthropic or OpenAI does not affect credits you already bought. The gateway exposes both OpenAI-compatible and Anthropic-compatible endpoints, which means it works with the official Python and TypeScript SDKs, LangChain, Claude Code, and most agent frameworks without code changes beyond the base URL. Every chat model accepts image and PDF input. Image generation through gpt-image-2 is also available, billed per image at three size tiers: $0.0177 for up to 1024 pixels square, $0.0264 for up to 2048 pixels square, and $0.0354 above that. What InferKey is not is a full-featured API management platform. There is no team workspace, no per-key spend cap, no provider failover, and no audit log beyond token counts and timestamps. The dashboard covers balance and per-request usage, and that is roughly the extent of the operational tooling. For a single developer who just wants cheaper tokens, that is fine. For anyone building a product with multiple contributors or compliance requirements, it is a real gap. The model lineup at the time of this review covers many Claude variants (Haiku, Sonnet, Opus, Fable) and GPT variants (5.4, 5.4-mini, 5.5, 5.6-luna, 5.6-terra, 5.6-sol, 6-luna), totaling 15 text models plus gpt-image-2 for image generation. InferKey is an independent gateway with no affiliation with Anthropic or OpenAI. ## Who should use it InferKey fits a solo developer or small indie project that already calls the Claude or GPT APIs directly and wants to cut that bill without rebuilding anything. The migration is genuinely minimal: one environment variable change for the base URL, one for the key. A freelancer running automated pipelines, a hobbyist building a personal assistant, or a developer prototyping an agent who wants to avoid a monthly subscription commitment will find the prepaid, no-expiry model convenient. Anyone building a production service with a team should look elsewhere. GitHub Copilot is the better choice if the primary use case is in-editor coding assistance rather than raw API access. If you need a more capable coding agent that can take a task end-to-end rather than just cheaper tokens, Cursor's integrated model access or Tura's local agent approach will serve you better than a bare API gateway. ## FAQ Q: Are the models the real Claude and GPT models, or fine-tuned versions? A: According to the pricing page, requests are relayed to the official models and responses are returned unmodified. InferKey acts as a pass-through relay, not a host of its own model weights. Q: How much does it cost to run a million tokens through claude-haiku-4-5? A: Input costs $0.51 per 1M tokens and output costs $2.55 per 1M tokens, compared to the official rates of $1 and $5 respectively, a 49% saving. Cache write is $0.63/1M and cache read is $0.06/1M. Q: Is there a free trial or starter credit? A: No. The pricing page lists no free tier and no trial credit. You must purchase credits before making any API calls. Q: What happens if my credits run out mid-request? A: Requests stop until you top up. The site states you can never overspend, and your balance plus per-request usage are visible in the dashboard in real time. Q: Does InferKey store my prompts? A: No. The FAQ on the pricing page states that requests stream through to the model without being stored. Only billing metadata is retained: token counts, model name, and timestamps. Q: Which payment methods are accepted? A: The pricing page lists card and Alipay as the two accepted payment methods. No cryptocurrency or invoice billing is mentioned.