image2.5 from $0.0085 per 1K image · Seedance 2.5 from $0.0961/sec · cached LLM input from $0.40/M — one OpenAI-compatible endpoint at https://api.apimart.ai/v1, no monthly plan required. (observed 2026-09-17)

Get an API key · Live pricing · Model page

Why teams route through APIMart

• One key, entire catalog. The same https://api.apimart.ai/v1 base URL and Authorization header reach the whole catalog behind one key and 300+ other image, video and language models — switch the model field, not your client.
• $1 minimum, pay as you go. No subscription and no prepaid plan to size up front: top up from $1 and spend it on calls. There is no free quota to burn through first, so the price in this table is the price you pay.
• The charge comes back in the response. Every call reports the amount billed (cost / credits_cost), so a spend number is read per call instead of guessed at month end.
• Async by design. Submit, take the task_id, poll GET /v1/tasks/{id} — batching and retries are ordinary queue work, not a bespoke integration.

LLM gateway comparison without the marketing table: pick the dimensions that matter to you, weight them, and score the four gateway archetypes (self-hosted open source, managed control plane, relayed per-unit route, direct vendor API) with a script you can rerun when your requirements change.

Vendor feature matrices go stale in weeks, and a comparison written by an interested party is not a comparison. This repository keeps the part that stays true: the dimensions on which gateways differ, plus a scoring script so the weighting is yours. Named products appear only as examples of an archetype.

python tools/compare.py --show                                  # the criteria matrix
python tools/compare.py --weight control=3 --weight cost_transparency=3 \
                        --weight setup_effort=2 --weight ops_burden=1

1. Freeze a request set. 20–50 real prompts covering every surface you use (chat, streaming, tools, images, vision).
2. Score the archetypes with tools/compare.py using weights your team agrees on — record who set which weight.
3. Test two candidates for real. Same prompts, same acceptance rules; record p50/p95 latency, error classes, retry behaviour and cost per accepted artefact.
4. Check the boring parts: key rotation, quota exhaustion behaviour, retention and region, and what happens to an in-flight job when you fail over.
5. Only then decide, and keep the rollback as a configuration change rather than a code change.

• It is not a ranking of vendors, and it does not claim that any archetype wins: a two-person team and a platform team with an SRE rotation should reach different conclusions from the same matrix.
• It is not a benchmark. Latency and price numbers belong in your own run, on your own prompts.

Linear at the observed per-unit rate, no volume discount assumed. Snapshot 2026-09-17; re-check the live table before committing a budget.

• llm gateway comparison
• ai api gateway
• llm gateway open source
• llm gateway vs proxy
• ai api aggregator
• openai compatible api
• ai api pricing comparison

Start with $1. Get an API key → check live pricing → open the whole catalog behind one key in the model library. The first call is three steps: submit, poll task_id, read the charged amount off the response.

Outbound APIMart links are minted through the promo link API; hand-made tracking parameters are rejected by tools/check_links.py in CI.

This repository describes gateway archetypes and a scoring method. It is published to document that method, not to claim official status for any vendor; product names appear as examples and belong to their owners. Verify each cell against the vendor's documentation for your own account.

README.md              comparison method, dimensions and workflow
data/gateways.json     archetypes, dimension values and scores (schema: llm-gateway-comparison-v1)
tools/compare.py       weighting and scoring
tools/check_links.py   attribution guard (CI)

MIT — see LICENSE.