{"rid":"compute/amd-radeon-cloud","card":{"rid":"compute/amd-radeon-cloud","eid":"amd-2","topic":"compute","slug":"amd-radeon-cloud","provider":"AMD Radeon Cloud","url":"https://developer.amd.com.cn/radeon/tokenfactory","domain":"amd.com.cn","category":"GPU cloud","tier":null,"verdict":"qualifies-with-caveat","vorder":1,"grp":"usable","confidence":"documented","binding":"request-rate","identity_kind":"any-phone","identity_basis":"stated","needs_card":0,"needs_phone":1,"needs_id":0,"needs_institution":0,"country":null,"crypto":null,"interfaces":"ssh,rest,cli","egress_metered":null,"per_object_bytes":null,"retention_days":null,"verified_at":"2026-10-22T00:00:00.000Z","verified_basis":"live","checked_on":"2026-10-22","stale_on":"2026-11-21","n_evidence":5,"usd_month":null,"usd_per_tb_month":null,"summary":"AMD runs two sites. On the China site the AMD AI Developer Program is free to join and new members start with 100 points, which is 100 GPU-hours at 1 point p...","duplicate_of":null,"n_unknown":9},"record":{"topic":"compute","slug":"amd-radeon-cloud","provider":"AMD Radeon Cloud","url":"https://developer.amd.com.cn/radeon/tokenfactory","category":"GPU cloud","free_tier":{"stated":"AMD runs two sites. On the China site the AMD AI Developer Program is free to join and new members start with 100 points, which is 100 GPU-hours at 1 point per hour; further points come from courses and activities, and an account may redeem at most 20 points a day. Token Factory, an OpenAI-compatible inference endpoint listing 16 public free models, is free, always on, and spends no credits.","unit":"hours","metered_on":"wall-clock","source_url":"https://developer.amd.com.cn/radeon/tokenfactory","read_on":"2026-10-22","notes":"one-time-credit"},"per_object_cap":{"stated":"not applicable","bytes":null,"source_url":"https://developer.amd.com.cn/radeon/tokenfactory","notes":null},"retention":{"stated":"not applicable","days":null,"source_url":"https://developer.amd.com.cn/radeon/tokenfactory","notes":null},"egress":{"terms":"not stated","metered":false,"free_allowance":null,"price_per_tb":null,"source_url":"https://developer.amd.com.cn/radeon/tokenfactory"},"interface":["ssh","rest","cli"],"interface_detail":"inferred from the tooling the provider documents; not probed","tooling":"not stated","chaining":"A recurring 20 GPU-hours a day on a ROCm container with SSH, a CLI and a persistent volume is the first free allowance found that is not a one-time credit or a notebook tab, and it composes with Token Factory: inference costs nothing, so the 20 hours can be spent entirely on training.","limits":[{"kind":"request-rate","value":"the 100 starting points buy 100 GPU-hours and redemptions are capped at 20 points a day, so the grant arrives in five daily slices; one account may hold one instance at a time and an instance bills while idle, so a forgotten container spends its credits in wall clock.","source":"https://developer.amd.com.cn/radeon/tokenfactory"}],"limits_note":"the 100 starting points buy 100 GPU-hours and redemptions are capped at 20 points a day, so the grant arrives in five daily slices; one account may hold one instance at a time and an instance bills while idle, so a forgotten container spends its credits in wall clock.","binding":"request-rate","binding_note":"the 100 starting points buy 100 GPU-hours and redemptions are capped at 20 points a day, so the grant arrives in five daily slices; one account may hold one instance at a time and an instance bills while idle, so a forgotten container spends its credits in wall clock.","identity_cost":{"kind":"any-phone","detail":null,"country":null},"reliability_note":"no reliability event recorded","confidence":"documented","evidence":[{"url":"https://amd-aim.github.io/radeon-cloud-docs/guides/developer-program/","date":"2026-10-22","proved":"quoted: the program is free and needs no AMD hardware; 'The program says new members start with 100 points'; '1 point buys 1 hour of compute on the China site'; 'Each account can exchange up to 20 points a day, so a single exchange is for 1 to 20 hours'"},{"url":"https://amd-aim.github.io/radeon-cloud-docs/guides/credits/","date":"2026-10-22","proved":"quoted: 'A running instance spends 1 credit per GPU every hour', the rate is shown per size as '1 GPU, 16 cores, 55GB, 1 credit / hour', an idle instance still spends credits, and the instance is deleted when credits run out"},{"url":"https://amd-aim.github.io/radeon-cloud-docs/resources/faq/","date":"2026-10-22","proved":"quoted: 'Each account can have one instance at a time. Destroy it to launch another.'; persistent /workspace is 25 GB and some accounts can choose 100 GB; a launch fails for no free node, insufficient credits, or an account still under review"},{"url":"https://developer.amd.com.cn/radeon/tokenfactory","date":"2026-10-22","proved":"read: 19 models total, 16 of them 'Public Free Model APIs' described as ready-to-use free API endpoints served by the AMD GPU Cloud, including DeepSeek-V4.1-Flash and DeepSeek-V4-Flash"},{"url":"https://linux.do/t/topic/2922778","date":"2026-09-19","proved":"the route that finds it, translated: a ModelScope notebook page offers an AMD GPU environment with 192 GB of VRAM after linking an AMD developer account, one user reports 8 hours maximum per boot and persistence only inside /workspace"}],"verdict":"qualifies-with-caveat","verified_at":"2026-10-22T00:00:00.000Z","notes":"The China site signs in with a mobile number or email plus a slider captcha, or with a ModelScope, CSDN or GitHub account, so the identity is a phone number and not a national ID; whether the AMD China account itself demands a real-name check after signup is unverified. The Global site at radeon-global.anruicloud.com signs in with email or GitHub and is to test first, because it needs no China number and its credit route is unstated in the docs read. Open question: does the Global site grant the same points, or none? The docs say the Global site does not use the developer program, which suggests it has no free credit at all - that asymmetry is the thing to settle.","extras":{"wave2":{"class":"vendor-developer-program","free_tier_kind":"one-time-credit","shape":{"vcpu":16,"ram_gib":55,"gpu_class":"AMD Radeon (ROCm)","gpu_count":1,"vram_gib":null}}}},"claims":[{"path":"$.free_tier.stated","kind":"free-quota","num":100,"unit":"GPU-hours","per":null,"currency":null,"usd":null,"fx_date":null,"parse":"ok"},{"path":"$.limits[0].value","kind":"limit:request-rate","num":100,"unit":"GPU-hours","per":null,"currency":null,"usd":null,"fx_date":null,"parse":"ok"}],"unknown":[["$.egress.metered","egress-unknown"],["$.egress.terms","not-stated"],["$.identity_cost.detail","process"],["$.per_object_cap.notes","process"],["$.per_object_cap.stated","not-applicable"],["$.reliability_note","none-observed"],["$.retention.notes","process"],["$.retention.stated","not-applicable"],["$.tooling","not-stated"]],"history":[],"drift":[]}