Side-by-side summary
| Groq | Cerebras | |
|---|---|---|
| Hardware | Groq LPU chips | Cerebras wafer-scale chips |
| Free access | Recurring Free plan with per-model limits | No renewing free tier. $5 trial credit for 30 days |
| Card or payment method | Not needed for the Free plan | Needed to activate the trial and the API |
| GPT-OSS 120B speed (vendor figure) | About 500 tokens/s | About 3,000 tokens/s |
| Qwen 3.8 27B speed (vendor figure) | About 450 tokens/s | About 1,850 tokens/s |
| GPT-OSS 120B free limit | See the live table below | Trial: 5 req/min, 1M tokens/day |
| Context on free access | 131,072 tokens on most chat models | 65K (GPT-OSS) and 64K (Qwen) on the trial |
| OpenAI-compatible API | Yes, api.groq.com/openai/v1 | Yes, api.cerebras.ai/v1 |
| Other services | Speech to text (Whisper) and text to speech | Chat models only in the current list |
Speed figures come from the model pages of each vendor. Real speed depends on prompt length, load and region. We did not measure them.
Speed
On the vendor figures, Cerebras is faster on the two models that both serve. It lists about 3,000 tokens per second for GPT-OSS 120B. Groq lists about 500 tokens per second for the same model. For Qwen 3.8 27B, Cerebras lists about 1,850 and Groq about 450.
Both are much faster than most GPU providers. For a chat reply of a few hundred tokens, both feel instant to a user. The difference matters more for long outputs, batch jobs and agents that chain many calls.
Groq also lists GPT-OSS 20B at about 1,000 tokens per second. If you need speed on a free plan, a small model on Groq is often the best choice.
Free limits
Groq applies its rate limits at the organization level. The table shows the Groq free limits in today's catalog.
| Model | Free on | Context | Free limits (best offer) | Tools | Vision |
|---|---|---|---|---|---|
| Qwen3.8 27B | Groq | 262K | 30 req/min 250 req/day 70,000 tokens/min ~6M | Yes | Yes |
| GPT-OSS 120B | Groq | 131K | 30 req/min 1,000 req/day 8,000 tokens/min 200,000 tokens/day ~6M | Yes | No |
| GPT-OSS 20B | Groq | 131K | 30 req/min 1,000 req/day 8,000 tokens/min 200,000 tokens/day ~6M | Yes | No |
| GPT-OSS Safeguard 20B | Groq | 131K | 30 req/min 1,000 req/day 8,000 tokens/min 200,000 tokens/day ~6M | Yes | No |
| ALLaM 2 7B | Groq | 4K | 30 req/min 1,000 req/day 6,000 tokens/min 500,000 tokens/day ~15M | No | No |
| Llama Prompt Guard 2 22M | Groq | 512 | 30 req/min 14,400 req/day 15,000 tokens/min 14.4K rpd | No | No |
| Llama Prompt Guard 2 86M | Groq | 512 | 30 req/min 14,400 req/day 15,000 tokens/min 14.4K rpd | No | No |
Live data from the FreeLLMAPI catalog (version 2026.11.19, checked 19 November 2026). Free limits change often. The catalog is checked every day.
Cerebras has no row in the free catalog, because its only free access is the trial. The trial gives $5 of credit after you add a verified payment method. The credit expires 30 days after Cerebras grants it. During the trial, GPT-OSS 120B allows 5 requests per minute and 1M tokens per day. Qwen 3.8 27B allows 300 requests per minute.
Models
- Groq production models: GPT-OSS 120B, GPT-OSS 20B, Llama 3.3 70B, Llama 3.1 8B and Whisper speech models.
- Groq preview models: Qwen 3.8 27B, MiniMax M2.7, GPT-OSS Safeguard 20B, Llama Prompt Guard and Orpheus speech models, with others.
- Cerebras models: GPT-OSS 120B and Qwen 3.8 27B.
- Preview models can change or go away with short notice. Do not build a product on a preview model only.
Both vendors serve GPT-OSS, so you can compare them on the same model. Other providers also serve GPT-OSS free. Read the free GPT-OSS API page for all routes.
Rate limit errors and headers
Both APIs return HTTP 429 when you go above a limit. Groq documents its rate limit headers. Read them to slow down before you reach the limit.
| Groq header | Meaning |
|---|---|
x-ratelimit-limit-requests | Your requests per day limit |
x-ratelimit-remaining-requests | Requests left in the current day |
x-ratelimit-limit-tokens | Your tokens per minute limit |
x-ratelimit-remaining-tokens | Tokens left in the current minute |
retry-after | Seconds to wait. Groq sends it only with a 429 response |
On the Cerebras trial, the most common limit is 5 requests per minute on GPT-OSS 120B. On Groq, the most common limit on long prompts is tokens per minute. For more detail, read Groq rate limit errors.
Which one to choose
| Use case | Choice | Reason |
|---|---|---|
| Free API for a side project | Groq | Recurring free plan with no card |
| Fastest output on GPT-OSS 120B | Cerebras | About 6 times the Groq vendor figure |
| Coding agent with long prompts | Groq, or a pool of providers | 131K context. The Cerebras trial context is about 65K |
| Speech to text | Groq | Groq serves Whisper models |
| A paid, very fast production endpoint | Cerebras | Developer tier has no hourly or daily caps |
| Free use that does not stop at one limit | Both inside a gateway | Failover moves requests when one limit is reached |
You do not have to pick one. FreeLLMAPI has adapters for both. Add a Groq key for free daily use and a Cerebras key during the trial. Use the auto:fast routing strategy to prefer fast models. When Groq returns 429, the router tries the next model in your chain.
Frequently asked questions
Is Groq faster than Cerebras?
No, not on the vendor figures. Cerebras lists about 3,000 tokens per second for GPT-OSS 120B. Groq lists about 500 tokens per second for the same model.
Is Cerebras free like Groq?
No. Groq has a recurring Free plan. Cerebras has no renewing free tier, only a $5 trial credit that expires after 30 days and needs a payment method.
What are the Groq free tier limits?
Groq sets limits per model, at the organization level. The table on this page shows the current free limits for each Groq model from the live catalog.
Which models do Groq and Cerebras both serve?
Both serve GPT-OSS 120B and Qwen 3.8 27B. Groq also serves GPT-OSS 20B, Llama models, Whisper and some preview models.
Can I use Groq and Cerebras together?
Yes. A gateway such as FreeLLMAPI can hold keys for both. When one provider returns a rate limit error, the gateway sends the request to the next provider.
Related guides
- Cerebras free tier: what you get in 2026
- Groq rate limit reached (429)
- free gpt oss api
- free groq api
- Free LLM API rate limits
- Free speech to text API
Sources
- GroqCloud docs: Supported models (checked 20 November 2026)
- GroqCloud docs: Rate limits (checked 20 November 2026)
- Groq Community: Is there a free tier and what are its limits? (checked 20 November 2026)
- Cerebras Inference docs: Rate limits (checked 20 November 2026)
- Cerebras Inference docs: Models overview (checked 20 November 2026)
- FreeLLMAPI docs: Supported platforms (checked 20 November 2026)