Amazon Bedrock vs GroqCloud

Amazon Bedrock 94

Model reading. No OpenAPI, but the bedrock-runtime and bedrock Smithy models are public in aws/api-models-aws with types, required members and enums, counted as for S3 (25). llms.txt for the user guide with about 2,000 lines, each page with a Markdown twin (10). Model cards state context, output limits, endpoints and supported functions per model, and the guide now recommends bedrock-runtime for new work, but support for each API, endpoint and feature changes model by model (16 of 20). Structured outputs compile JSON Schema to a grammar, with minimum, maximum and string lengths supported, though some inference profiles don't support it, such as Haiku 4.5 through the India profile (13 of 15). Every operation lists typed errors with HTTP codes and links to a troubleshooting page with causes and fixes, and examples run through the guide (15). API version 2023-09-30 and a dated document history with an RSS feed, newest entry 30 October 2026, with structured outputs for Claude on both endpoints added on 18 October (15).

GroqCloud 64

No OpenAPI document in the docs, and the SDK's .stats.yml carries only an endpoint count of 17, no spec URL (0). llms.txt at console.groq.com/llms.txt (10). Reference not read in full this run (15 of 20). Structured outputs have a Strict mode as well as Best-effort (15). The errors page lists 15 status codes with recovery advice, including 498 for Flex capacity and 424 for remote MCP auth, and the error object with message and type (14 of 15). The changelog at console.groq.com/docs/changelog, read on 1 November, is dated, but its newest entry is 18 April 2026 and it doesn't record the four shutdowns since July, which only the deprecations page carries (10 of 15).

Amazon Bedrock 98

Model reading. IAM with SigV4, roles and STS session credentials, policies down to a model or inference-profile ARN, short-term API keys that expire within 12 hours, long-term keys with expiry and reset, and condition keys that can deny bearer tokens altogether (30). The security page says customer data isn't shared with model providers or used to improve the base models (20). The abuse detection page says Bedrock stores no inputs or outputs by default, and a per-Region none mode blocks models that require retention, but Claude Fable 5 and 5.1 keep all traffic up to 30 days and GPT-6 and GPT-5.x keep flagged traffic up to 30 days (13 of 15, a judgement call; Anthropic's own API got 12 for the same Fable exception without an account-level block). CloudTrail logs API calls, and model invocation logging can send prompts and responses to S3 or CloudWatch Logs (15). Bedrock is in ISO, SOC and CSA STAR Level 2 scope, HIPAA eligible and FedRAMP High in GovCloud (US-West) per its security page, with a vulnerability disclosure programme on HackerOne, and the aws.amazon.com security.txt, expired on 6 October, was renewed with Expires 2027-10-16 when read on 6 November (20). Up from 96 on 6 October for that reason.

GroqCloud 82

Model reading. Bearer keys scoped to a project, with custom request limits per project and model permissions at organisation and project level. The security onboarding page advises rotating keys periodically and revoking them at once if compromise is suspected (30). The services agreement, read first-hand on 6 October and not re-read on 17 October, says Groq is not permitted to use inputs or outputs for training unless the customer grants permission (20). No retention by default, up to 30 days for reliability and abuse monitoring, and zero retention as a setting any customer can turn on in Data Controls, per the data page (15). Request logs, usage per project and a read-only Reader role (12 of 15). groq.com's security.txt holds a Contact line and nothing else, and the trust centre at trust.groq.com is a Vanta page that renders only with JavaScript, so certifications and any bug bounty are unchecked (5 of 20).

Amazon Bedrock 92

Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Converse takes tools with toolChoice auto, any or a named tool and strict: true, but forced choice depends on the model behind it (17 of 20). Structured outputs on supported models only (13 of 15). Explicit and implicit prompt caching, cache reads at $0.20 per million on Opus 5.5, 0.05x input, with 5-minute and 1-hour TTLs (15). 1M-token context on Opus 5.5, Sonnet 5.5 and Nova 2 Lite, a judgement call since Bedrock has no default model (15). Batch at half price on supported models, but the pricing page shows none for Opus 5.5 or Sonnet 5.5, while Flex is half price on Nova (7 of 10). AWS SDKs in every major language plus token generators for Python, JavaScript and Java (10). Typed errors with recovery advice per code (15).

GroqCloud 80

Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Tool use documented, parallel calls and forced choice not checked (15 of 20). Strict structured outputs (15). Automatic prompt caching at 50% off cached input, with a 2-hour cache life, on gpt-oss-20b, gpt-oss-120b and gpt-oss-safeguard-20b only (10 of 15). 131,072-token context on every self-serve model; only MiniMax M2.7, a preview on enterprise pricing, reaches 196,608 (5 of 15). Batch at half price (10). Official SDKs for Python and JavaScript (10). Errors page with codes, recovery advice and a typed error object (15).

Amazon Bedrock 80

AWS Health Dashboard with per-service, per-Region feeds and a public history file (20). The history lists Amazon Bedrock as disrupted in ME-CENTRAL-1 (UAE) and ME-SOUTH-1 (Bahrain) since March 2026, where AWS said on 15 September 2026 (still inside the 90-day window on 6 November) that it can't restore resources held only there. No other Bedrock event appears in the history in the 90 days to 6 November; an EU-SOUTH-2 packet-loss event on 4 October named Bedrock AgentCore, not Bedrock. This counts as one major event (10). StatusGator's third-party mirror shows 'Increased Error Rates' warnings on 24, 26 and 27 August and 10 September that we couldn't find in AWS's own history, so we didn't score them. Per-model token quotas with default values in the AWS General Reference, such as 30,000,000 tokens a minute for Opus 5.5 global inference (15). The error guide gives exponential backoff with jitter for 429, 503 and 529 and says to honour Retry-After, and the SDKs retry ModelNotReadyException up to five times (15). SLA of 99.9 per cent a month per Region for the model APIs, last updated 4 October 2023 (10). Converse and the current Claude, GPT-6 and Nova 2 Lite models are GA (10).

GroqCloud 90

Status page at groqstatus.com on incident.io, with components for the API, each production and preview model and the website (20). The page's RSS feed and the incident pages it links, read on 1 November, list three incidents in the 90-day window, all in October, and nothing since 12 October apart from a console maintenance on 18 October. On 10 October the enterprise-only preview MiniMax M2.7 had failed requests from data centre hardware issues from 10:09 to 16:14 UTC. The incident's header and impact record say degraded performance, while its Identified update set the model to partial outage and warned of failed requests. On 12 October console.groq.com returned 500 errors on metrics, /models and key creation from 03:17 to 03:38, and the vendor said inference requests were not affected. On 12 October MiniMax M2.7 had a partial outage from 18:17 to 18:36. None took down a core API for an hour, since the six-hour incident hit one enterprise preview model that the status page describes as for evaluation only, so minor incidents only (20). A stricter reading would count it as a major and give 10. Limits published per model, 30 requests a minute, 1,000 a day and 8,000 tokens a minute on gpt-oss for the free plan, re-read on 1 November (15). 429 carries retry-after, x-ratelimit-* headers come on every response, and the errors page says to wait before retrying a 503 and that a 502 may clear on retry (15). The Performance Tier, enterprise only, lists a 99.9% availability SLA with details in the offline agreement, per the 6 October check, not re-read on 1 November (10). gpt-oss and Whisper models are production, Qwen 3.8 27B, MiniMax M2.7 and Muse Spark Open are previews (10). 20+20+15+15+10+10 = 90.

Amazon Bedrock 30

No x402, MPP or L402 (0). Per-token prices per model, tier and Region on the public pricing page, filled from AWS's public price feed (20). New accounts get up to $200 in Free Tier credits and the Free Tier page lists Bedrock among eligible services, and the Free Tier FAQ says most new customers don't need a payment method, but AWS may still ask for one, so half, as for S3 (10). A person signs up for AWS in a browser and sets up IAM before any key exists (0).

GroqCloud 40

No machine payment protocol (0). Per-token prices on the models page, read without a login, such as gpt-oss-120b at $0.15 in and $0.60 out per million (20). Free plan with no card (20). Console sign-up in a browser (0).

Amazon Bedrock 81

Model reading. The last model or API change is 30 October 2026 (OpenAI model API support and cross-Region inference in GovCloud), after structured outputs for Claude on both endpoints on 18 October (30). Models launched since 7 September 2026 get a 6-month or 45-day Legacy period, but the legacy page, which promises at least 6 months in Legacy, lists Claude 3 Haiku, 3 Sonnet, 3.5 Sonnet, 3.5 Sonnet v2 and 3.7 Sonnet as Legacy from 10 or 13 October with end of life on 13 November in several Regions, about 34 days (6 of 12, down from 10). Six models reached end of life in the 90 days to 6 November on five dates, Command R and R+ on 19 August, Claude 3 Haiku on 10 September, Nova Sonic on 14 September, Nova Canvas and Nova Reel on 30 September and Claude Sonnet 4 on 14 October (3 of 8, down from 4). Dated document history several times a month, What's New and re:Post, with direct support on a paid plan (12 of 15). SDK issue replies not sampled (5 of 10). Current official SDKs, boto3 1.43.125 on 5 November and @aws-sdk/client-bedrock-runtime 3.1163.0 (15). Supported runtimes stated, Python 3.10 or later and Node 20 or later (10). Down from 86 on 6 October.

GroqCloud 72

Model reading. Meta's Muse Spark Open went into limited preview on GroqCloud on 15 October 2026, per Groq's announcement, 17 days before this check, and the models page lists it as an enterprise model (30). The newest shutdown is still groq/compound and compound-mini on 21 September. The changelog's newest entry is 18 April 2026, the Python SDK 1.7.0 (tagged 25 August) and TypeScript SDK 1.6.0 are still the newest tags, with commits to 4 September and no release since, and the deprecations page has no announcement after 24 August. Production models get an email and a migration path, previews may go at short notice, and no minimum period is stated. Llama 3.1 8B and 3.3 70B got 60 days on the free and developer tiers, Compound 28 days with no replacement named (4 of 12). Three shutdown dates in the 90 days to 1 November, 16 August, 14 September and 21 September. Three in 90 days is still frequent churn (0 of 8). A dated changelog with an RSS feed and a community forum exist, but the changelog's newest entry is 18 April 2026; support replies not checked (10 of 15). SDK issue replies not sampled, since GitHub's API refused our shell (5 of 10). Python SDK 1.7.0 and TypeScript SDK 1.6.0, both published on 26 August 2026 and still the newest tags on 1 November (15). CI, release-please and lock-file vulnerability fixes in 1.7.0 (8 of 10). 30+4+0+10+5+15+8 = 72.