I’m experiencing a 403 Forbidden error when using Vercel AI Gateway, even though my account has a valid payment method attached and I still have free AI credits remaining.
The API key appears to be valid because the /v1/models endpoint works, but /v1/chat/completions is blocked.
I’m experiencing the exact same issue. Since yesterday (May 24, 2026), most of the AI Gateway models suddenly stopped working when using free credits. Previously, this restriction was only applied to flagship models like Claude Opus and Gemini Pro, but now it seems models like Sonnet, DeepSeek and GLM are also restricted.
Could the team clarify if this is just a temporary measure or a permanent policy change?
I am writing to report that I am experiencing the same issue. It appears that the majority of models are currently unavailable through the AI Gateway.
While I understand that flagship models may be restricted for users on free credits, the outage currently extends to most models, including Kimi, DeepSeek, GLM, and Grok. This persists even for the Lite/Flash versions of these models.
Additionally, models that are still available are getting heavily rate limited.
Could you please clarify if this is a known issue and if it is truly a temporary measure?
I also met the same issue when using deepseek-v4-flash.
{“error”:{“message”:“Free tier users do not have access to this model. Upgrade to paid credits at Vercel for unrestricted access.”,“type”:“no_providers_available”,“param”:{“error”:“Free tier users do not have access to this model. Upgrade to paid credits at Vercel for unrestricted access.”,“type”:“no_providers_available”,“statusCode”:403,“name”:“RestrictedModelsError”,“message”:“Free tier users do not have access to this model. Upgrade to paid credits at Vercel for unrestricted access.”}},“providerMetadata”:{“gateway”:{“routing”:{“originalModelId”:“deepseek/deepseek-v4-flash”,“resolvedProvider”:“deepseek”,“fallbacksAvailable”:[“novita”,“deepinfra”],“canonicalSlug”:“deepseek/deepseek-v4-flash”,“modelAttemptCount”:1,“modelAttempts”:[{“canonicalSlug”:“deepseek/deepseek-v4-flash”,“success”:false,“providerAttemptCount”:0,“providerAttempts”:}],“totalProviderAttemptCount”:0},“generationId”:“gen_01KSRSQ4F7ZTN38C6GJCX3FKAA”}}}
Same issue as wowpala. Either the documentation is wrong in implying that the deepseek-v4-flash model will use free credits, or the error is the source of truth and it’s a restricted paid-only model (for some reason).
Which is it? If documentation is outdated, then it should be updated, clearly.
The strange thing though - I added a card to file, several in fact. I tried busting cache - it still doesn’t work. So the error message it notes about not having a payment method is misleading, and even having a payment method doesn’t resolve it - it’s simply restricted to paid pro version.
I have a hunch this might also be related to rate limiting. When I get rated limited on other models, suddenly models that worked previously (such as deepseek-v3.2) start claiming that they are not available to free tiers. I think it’s more likely that they’re just failing for a generic “502” and are actually rate-limited.
EDIT: This seems to be the case. I waited a while and tested again. Model worked for a moment - both those with rate limit and the ones that said they were unavailable to free tiers. However, when I hit the rate limit again, the messages returned.
I’m experiencing the same issue and wanted to share the response I received from the support team.
The Problem: When trying to call models like Gemini and gpt-image-2 with only free credits, I receive the error: “Free tier users do not have access to this model.”
Support Team Response: The Vercel support team provided the following explanation:
“Some high-performance models and specific models in AI Gateway are restricted to prevent abuse and ensure safety. These models are not accessible with free tier credits alone.”
The Core Issue: However, the official documentation at AI Gateway Pricing still displays “All available models” for Free tier Model access. This does not reflect the actual behavior.
Request:
Please clearly document in the official documentation which models are not available on the free tier
Or, please resolve these restrictions as soon as possible
A prompt update to the documentation would be greatly appreciated for clarity.
I have the same issue with deepseek-v4-flash, free tier. Then when I login to Vercel AI Gateway dashboard to test the model from Vercel directly, the Add payment alert pops up, despite I already have credit on file.
I see many people having the same problem, so I wanted to ask. I get proprietary models were limited to stop abuse. But now it looks like the same limits are also on open-weight models. The models that are still available have very tight rate limits. It is hard to use them.
I checked the AI Gateway docs, but I cannot find any detailed mention of these rate limits or which specific models are affected. Is this really a temporary change? Any information would help a lot.
