Cerebras Inference
cerebras.ai · 4 records in Coding subscriptions, Compute, Inference · JSON
Records 4
- trialCerebras Code
Coding subscriptionsno free tieridentity cardlimit volume-quotaevidence documentedstale on 2026-12-04
- trialCerebras Inference
Computefree tieridentity cardlimit retention-clampedevidence documentedstale on 2026-12-05
- trialCerebras
Inferenceidentity cardlimit retention-clampedevidence documentedstale on 2026-11-28
- fails: identityCerebras Inference Research
Computefree tieridentity institutionno cardlimit write-gatedevidence documentedstale on 2026-11-22
Cerebras Code
Coding subscriptions- Free allowance
- No free Code plan and none under USD 20: Pro is USD 50 and Max USD 200, both marked 'sold out'.
- Verdict
- triala time-boxed or one-time trial, not a recurring allowance
- Cost
- no price stated
- Identity
- cardcard required, no phone
- Binding limit
- volume-quotaPro: up to 24 million tokens per day; Max: up to 120 million tokens per day; Free Trial: 5 requests a minute, 30K uncached tokens a minute, 1M tokens an hour and 1M tokens a day per model, on gpt-oss-120b and qwen-3.8-27b
- Evidence
- documented27 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-12-04checked 2026-11-03 (live); stale on 2026-12-04
| Free tier | No free Code plan and none under USD 20: Pro is USD 50 and Max USD 200, both marked 'sold out'. The page's 'Free Trial' is 'Limited Free Credits', the platform trial: a one-time USD 5 promotional credit after a valid payment method is added, expiring 30 days after activation unit credits; metered on unstated; the trial credit is spent per token at the Free Trial rate limits; the Code plans meter tokens per day |
|---|---|
| Price | none stated |
| Limits |
|
| Binding | volume-quota with Pro and Max sold out, the reachable route is the one-time trial at 1M tokens a day per model, which ends when the USD 5 credit is spent or 30 days pass |
| Identity | card a cloud.cerebras.ai account and a valid payment method for the trial credit: 'Add a valid payment method to receive a one-time $5 promotional credit'; Code Pro and Max are prepaid monthly subscriptions |
| Egress | not applicable |
| Interface | rest an API key from cloud.cerebras.ai used from any editor or agent that accepts an OpenAI-compatible endpoint; the vendor names Cline, OpenCode and Crush. There is no Cerebras agent, so the headless mode is whichever agent calls the key |
| Tooling | any OpenAI-compatible coding agent or SDK |
| Composes with | a key for Cline, OpenCode, Crush or Aider; a support article names a referral count that raises an account's limits to the Code Pro equivalent and states no duration; the current referral terms instead grant 200,000 bonus tokens per day to referrer and referee, capped at 1,000,000 tokens total per user and at five successful referrals, so the two pages do not agree and neither says how long a grant lasts (unverified: behind a sign-up, the account's Limits view) |
| Reliability | both Code plans are marked sold out; the archived FAQ states operation at near 100% utilisation with requests briefly queued at peak, and 429 errors from per-second limits during bursts |
| Notes | Not a plan at a low price: the cheapest Code plan is Pro and both Pro and Max are sold out; whether either can be bought at all is unverified: behind a sign-up (the sold out buttons link to cloud.cerebras.ai). Model: cerebras.ai/code names GLM 4.7 for Pro, Max and the Free Trial on 2026-11-02, while the shared model catalog lists only gpt-oss-120b and qwen-3.8-27b, the docs state the Free Trial reaches every model of that catalog, and the docs deprecation list records zai-glm-4.7 as deprecated on 2026-08-17; no public page other than the product page names the model a Code key serves, and the key's own model list is unverified: behind a sign-up. The referral route: how long a Code Pro limit granted by referrals lasts is stated on no public page (unverified). |
| Evidence note | documented prices, sold-out state, daily token allowances, trial terms and the referral threshold are quoted from first-party pages read on 2026-10-30; the per-minute plan limits come from an archived copy of a FAQ that now answers 404. Whether a sold-out plan can be bought at all is shown only after sign-in. |
| Parsed numbers |
|
Evidence 27
| date | source | what it proved |
|---|---|---|
| 2026-10-30 | www.cerebras.ai/code | Pro at USD 50 serves GLM 4.7 with up to 24 million tokens a day |
| 2026-10-30 | www.cerebras.ai/code | Max at USD 200 allows up to 120 million tokens a day |
| 2026-10-30 | www.cerebras.ai/code | the Pro and Max buttons both read sold out |
| 2026-10-30 | www.cerebras.ai/code | the page's Free Trial is a limited-credit trial, not a recurring plan |
| 2026-10-30 | www.cerebras.ai/code | the plan is an API key for third-party agents |
| 2026-10-30 | www.cerebras.ai/pricing | the trial needs a payment method and grants USD 5 once, expiring after 30 days |
| 2026-10-30 | inference-docs.cerebras.ai/support/rate-limits | Free Trial limits are 5 RPM, 30K uncached TPM, 1M tokens an hour and 1M a day per model |
| 2026-10-30 | inference-docs.cerebras.ai/support/rate-limits | there is no recurring free allowance |
| 2026-10-30 | inference-docs.cerebras.ai/models/overview | the shared catalog, which the trial reaches, lists only gpt-oss-120b and qwen-3.8-27b; GLM 4.7 is absent from it |
| 2026-10-30 | web.archive.org/web/20260422081637/https://support.cerebras. | Code Pro allowed 50 requests and 1,000,000 tokens a minute and 24M tokens a day, as archived on 2026-04-22 |
| 2026-10-30 | web.archive.org/web/20260422081637/https://support.cerebras. | Code Max allowed 120 requests and 1,500,000 tokens a minute and 120M tokens a day, as archived on 2026-04-22 |
| 2026-10-30 | web.archive.org/web/20260422081637/https://support.cerebras. | the vendor reported queueing at peak load |
| 2026-10-30 | support.cerebras.net/articles/9996007307-cerebras-code-faq | the live Code FAQ URL answers 404, so the per-minute plan limits rest on the archived copy |
| 2026-10-30 | support.cerebras.net/articles/9150857851-how-to-get-your-fre | 11 referred sign-ups raise an account's limits to the Code Pro equivalent |
| 2026-10-30 | www.cerebras.ai/blog/introducing-cerebras-code | the plans launched on Qwen3-Coder at the same USD 50 and USD 200 prices; the model has since changed to GLM 4.7 |
| 2026-10-30 | tomrochette.com/agents/model-access/cerebras-code/ | a third party records both plans as sold out on 2026-10-14 and earlier |
| 2026-11-02 | www.cerebras.ai/referral-program | the referral terms cap total bonus accrual at 1,000,000 tokens per user |
| 2026-11-02 | inference-docs.cerebras.ai/support/deprecation | the docs deprecation list records zai-glm-4.7 as deprecated (entry dated 2026-08-17) |
| 2026-11-02 | inference-docs.cerebras.ai/support/deprecation | the deprecation entry for zai-glm-4.7 is dated 2026-08-17 |
| 2026-11-02 | inference-docs.cerebras.ai/models/overview | the shared catalog page re-read on 2026-11-02 lists only gpt-oss-120b and qwen-3.8-27b |
| 2026-11-02 | inference-docs.cerebras.ai/support/rate-limits | the Free Trial reaches exactly the shared catalog models, not GLM 4.7 |
| 2026-11-02 | www.cerebras.ai/referral-program | the referral terms cap bonus-earning referrals at five |
| 2026-11-02 | www.cerebras.ai/referral-program | each referral grants 200,000 bonus tokens a day to both parties |
| 2026-11-02 | www.cerebras.ai/referral-program | referral eligibility is limited to the United States, with an addendum for listed regions |
| 2026-11-02 | www.cerebras.ai/referral-program-addendum | the international addendum lists the eligible regions outside the United States |
| 2026-11-02 | inference-docs.cerebras.ai/support/rate-limits | the Free Trial credit needs a verified payment method |
| 2026-11-02 | www.cerebras.ai/code | Pro and Max both still read sold out on 2026-11-02 |
Unknown values 3
- egress terms: the field does not apply to this kind of offer
- per object cap stated: the field does not apply to this kind of offer
- retention stated: the field does not apply to this kind of offer
Cerebras Inference
Compute- Free allowance
- No recurring free allowance.
- Verdict
- triala time-boxed or one-time trial, not a recurring allowance
- Cost
- free tier
- Identity
- cardcard required, no phone
- Binding limit
- retention-clampedthe 30-day credit expiry binds before any rate does: USD 5 at gpt-oss-120b paid output pricing of USD 0.75 per 1M tokens is about 6.7M output tokens in total, or about 3.8M at a 3:1 input-to-output mix, so the whole allowance is one...
- Evidence
- documented6 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-12-05checked 2026-11-04 (live); stale on 2026-12-05
| Free tier | No recurring free allowance. The FAQ reads: 'Is there a permanently free tier? No. The Free Trial is time- and credit-bounded: $5 in credits that expire 30 days after they are granted.' The Free Trial row that carries the eye-catching numbers (gpt-oss-120b and qwen-3.8-27b at 5 RPM, 30K uncached TPM, 90K total TPM, 1M TPH, 1M TPD) sits inside that 30-day, card-gated window, and the same page states 'New accounts receive $5 in free credits after adding a verified payment method.' The trap is that the per-model pages have not caught up: the gpt-oss-120b model page still prints 'Free tier rate limits: 30 requests/min, 60k input tokens/min, 1M tokens/day' as though a free tier existed. unit requests; metered on wall-clock; trial-credit |
|---|---|
| Price | none stated |
| Limits |
|
| Identity | card card: a payment method is mandatory before any access - 'If you skip adding a payment method at sign-up, Playground and API access remain inactive until you do' - and the USD 5 arrives only 'after adding a verified payment method'. The pricing footnote adds 'You will not be charged or automatically enrolled in paid usage', so no authorisation or hold amount is stated in the evidence. Whether debit or prepaid is accepted is not stated: unknown. |
| Egress | not stated |
| Retention | the 30-day credit expiry binds before any rate does: USD 5 at gpt-oss-120b paid output pricing of USD 0.75 per 1M tokens is about 6.7M output tokens in total, or about 3.8M at a 3:1 input-to-output mix, so the whole allowance is one medium batch and not a monthly budget. |
| Interface | rest inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | nothing composes into a recurring job: the 1M-tokens-a-day ceiling is real for 30 days and then access pauses until credits are purchased. |
| Reliability | no reliability event recorded |
| Notes | Corrects the shape of a claim that circulates widely: '1M tokens a day free on Cerebras'. The number is a daily rate limit inside a card-gated 30-day trial, so the monthly figure is not 30M tokens. Arithmetic, labelled estimate: USD 5 / (USD 0.75 per 1M output) = 6.67M output tokens if the credit were spent on output alone; at a 3:1 input-to-output split, USD 5 buys about 3.8M tokens of mixed traffic. The contradiction between the FAQ and the model pages is the real finding for a future reader: a session that reads only the model page records a free tier that the FAQ denies in as many words. See also the SambaNova record in a10-repos, the same pattern of a stale sub-page describing a withdrawn free tier. |
| Parsed numbers |
|
Evidence 6
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | inference-docs.cerebras.ai/support/rate-limits | FAQ answer 'Is there a permanently free tier? No. The Free Trial is time- and credit-bounded: $5 in credits that expire 30 days after they are granted', quoted |
| 2026-10-22 | inference-docs.cerebras.ai/support/rate-limits | FAQ answer 'New accounts receive $5 in free credits after adding a verified payment method. These credits expire 30 days after they are granted', quoted; and 'If you skip adding a payment method at sign-up, Playground and API access remain inactive until you do' |
| 2026-10-22 | inference-docs.cerebras.ai/support/rate-limits | Free Trial table: gpt-oss-120b and qwen-3.8-27b at 5 RPM, 30K uncached TPM, 90K total TPM, 1M TPH, 1M TPD, quoted |
| 2026-10-22 | www.cerebras.ai/pricing | pricing page footnote '*** Add a valid payment method to receive a one-time $5 promotional credit. The credit expires 30 days after activation. You will not be charged or automatically enrolled in paid usage.', quoted |
| 2026-10-22 | inference-docs.cerebras.ai/models/openai-oss | model page still prints 'Free tier rate limits: 30 requests/min, 60k input tokens/min, 1M tokens/day' alongside paid pricing of USD 0.35 input and USD 0.75 output per 1M tokens, which contradicts the FAQ's 'there is no permanently free tier' |
| 2026-11-03 | inference-docs.cerebras.ai/api-reference/chat-completions | the API is an HTTP REST endpoint at api.cerebras.ai |
More fields 1
| facets | {"class": "refutation","free_tier_kind": "trial-credit","shape": {"vcpu": null,"ram_gib": null,"gpu_class": "wafer-scale WSE","gpu_count": null}} |
|---|
Unknown values 5
- egress metered: egress metering was not published
- egress terms: a source was read and is silent on this field
- per object cap stated: the field does not apply to this kind of offer
- reliability note: nothing of this kind was observed
- tooling: a source was read and is silent on this field
Cerebras
Inference- Free allowance
- A Free Trial, not a free tier: USD 5 of credits granted after a verified payment method is added, expiring 30 days after they are granted.
- Verdict
- triala time-boxed or one-time trial, not a recurring allowance
- Cost
- not stated
- Identity
- cardcard required, no phone
- Binding limit
- retention-clampedthe USD 5 of trial credits expire 30 days after they are granted, and API access stops when they run out or expire
- Evidence
- documented7 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-11-28checked 2026-10-28 (live); stale on 2026-11-28
| Free tier | a Free Trial, not a free tier: USD 5 of credits granted after a verified payment method is added, expiring 30 days after they are granted. The page answers 'Is there a permanently free tier?' with 'No.' unit none; metered on request-count; the trial's per-model limits are published: gpt-oss-120b and qwen-3.8-27b each at 5 RPM, 30K uncached TPM, 90K total TPM, 1M TPH and 1M TPD |
|---|---|
| Price | none stated |
| Limits |
the Developer (Pay as You Go) tier removes the hourly and daily token caps and raises the per-model rates: gpt-oss-120b 1M uncached TPM, 3M total TPM and 1K RPM; qwen-3.8-27b 150K uncached TPM, 750K total TPM and 300 RPM; that tier is pay-as-you-go and not free |
| Binding | retention-clamped the 30-day credit expiry binds before any rate does: the whole allowance is USD 5 that disappears, so there is no recurring quota to design against |
| Identity | card a verified payment method is required before the USD 5 of trial credits is granted; without a payment method the Playground and API access remain inactive |
| Egress | not published |
| Retention | the trial credits expire 30 days after they are granted (30 days) |
| Interface | rest an OpenAI-compatible chat completions endpoint at https://api.cerebras.ai/v1 |
| Tooling | any OpenAI-compatible client; curl against POST {base}/chat/completions |
| Composes with | pairs with an agent only inside a 30-day trial and only after a card; the scarce link is the verified payment method, which turns a free offer into a payment relationship |
| Reliability | an unauthenticated call returns 401 with 'Wrong API Key' and a 99-byte body, so an unauthenticated probe cannot distinguish a live endpoint from an expired one |
| Notes | The vendor's own page says there is no permanently free tier: the offer is USD 5 of credits that expire in 30 days and require a verified payment method. The identity cost is therefore card, not email, and the verdict is trial-only. |
| Evidence note | documented the trial's terms, its expiry, its per-model rates and the vendor's explicit 'No' to a permanently free tier are all quoted from the rate-limits page read 2026-10-22 at 200 with 393345 bytes. The 401 gate was also re-probed at 2026-10-22T04:42:43Z. |
| Parsed numbers |
|
Evidence 7
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | inference-docs.cerebras.ai/support/rate-limits | 'New accounts receive $5 in free credits after adding a verified payment method. These credits expire 30 days after they're granted'; 'Is there a permanently free tier? No. The Free Trial is time- and credit-bounded'; and the Free Trial table giving gpt-oss-120b 5 RPM, 30K uncached TPM, 90K total TPM, 1M TPH and 1M TPD |
| 2026-10-22 | probe: https://api.cerebras.ai/v1/chat/completions | an unauthenticated POST returned 401 with a 99-byte body, {"message":"Wrong API Key","type":"invalid_request_error","param":"api_key"}. The endpoint exists and gates on a key; no key existed here |
| 2026-10-28 | inference-docs.cerebras.ai/support/rate-limits | the Free Trial table lists qwen-3.8-27b at 5 RPM, 30K uncached TPM, 90K total TPM, 1M TPH and 1M TPD (gpt-oss-120b has the same row) |
| 2026-10-28 | inference-docs.cerebras.ai/support/rate-limits | the Developer (Pay as You Go) table lists qwen-3.8-27b at 150K uncached TPM, 750K total TPM and 300 RPM |
| 2026-10-28 | inference-docs.cerebras.ai/support/rate-limits | API access stays inactive until a payment method is added |
| 2026-10-28 | inference-docs.cerebras.ai/support/rate-limits | the Free Trial is USD 5 of credits granted after a verified payment method is added and expiring 30 days after they are granted |
| 2026-10-28 | inference-docs.cerebras.ai/support/rate-limits | the vendor answers 'Is there a permanently free tier?' with No: the trial is USD 5 of credits that expire 30 days after they are granted |
Unknown values 2
- egress metered: egress metering was not published
- per object cap stated: the field does not apply to this kind of offer
Cerebras Inference Research
Compute- Free allowance
- 'Selected Principal Investigators (PIs) may receive the following: Cerebras Inference service credits, equivalent to no more than $50,000 USD of inference credits' plus 'tutorials and hands-on sessions with Cerebras engineers and researchers'.
- Verdict
- fails: identitythe identity required is not reachable without it
- Cost
- free tier
- Identity
- institutionno card, no phone
- Binding limit
- write-gateda written proposal reviewed by Cerebras, with a principal investigator attached to a university, is the gate; the currency is inference tokens rather than machine time, so no node-hour figure exists to plan against.
- Evidence
- documented3 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-11-22checked 2026-10-22 (live); stale on 2026-11-22
| Free tier | 'Selected Principal Investigators (PIs) may receive the following: Cerebras Inference service credits, equivalent to no more than $50,000 USD of inference credits' plus 'tutorials and hands-on sessions with Cerebras engineers and researchers'. The call is addressed to 'university faculty and researchers'. The amount is capped by the vendor and 'The final grant amount and details will be determined by Cerebras.' unit none; metered on wall-clock; grant-credits |
|---|---|
| Price | none stated |
| Limits |
|
| Identity | institution not published |
| Egress | not stated |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | It would compose with the free Cerebras inference tier already recorded, because both spend one account. It does not compose for an individual: the PI requirement is the wall, and the USD 50,000 is the largest single number that a person without an affiliation cannot touch. |
| Reliability | no reliability event recorded |
| Notes | The rung is documented, and the number is deliberately recorded with its ceiling wording. What is unknown and left null: the application window, whether the RFP is still open, and the token-to-dollar rate used to denominate the award. A reader should treat this as a closed door with a big number behind it, not as capacity. |
| Parsed numbers |
|
Evidence 3
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | www.cerebras.ai/blog/grantfrp | 'Cerebras Inference service credits, equivalent to no more than $50,000 USD of inference credits', quoted, in the Grant section |
| 2026-10-22 | www.cerebras.ai/blog/grantfrp | 'We invite university faculty and researchers to respond to this Request for Proposals (RFP)', quoted |
| 2026-10-22 | www.cerebras.ai/blog/grantfrp | 'The final grant amount and details will be determined by Cerebras', quoted, so the USD 50,000 is a ceiling and not an award |
More fields 1
| facets | {"class": "institution-grant","free_tier_kind": "grant-credits","shape": {"vcpu": null,"ram_gib": null,"gpu_class": "Cerebras Wafer Scale Engine 3, inference only","gpu_count": null}} |
|---|
Unknown values 7
- egress metered: egress metering was not published
- egress terms: a source was read and is silent on this field
- interface: the access interface was not established
- per object cap stated: the field does not apply to this kind of offer
- reliability note: nothing of this kind was observed
- retention stated: the field does not apply to this kind of offer
- tooling: a source was read and is silent on this field