Amazon Bedrock vs OpenAI API

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).

OpenAI API 100

Official OpenAPI document (openapi.yaml and openapi.json) in the openai/openai-openapi repository, whose openapi.yaml (version 2.3.0, updated 31 October) lists gpt-6.1-terra (25). llms.txt index with per-product files, and every docs page has a Markdown twin at .md (10). The reference documents each parameter, and model pages say which model fits which job (20). Structured outputs with strict: true on schemas and function tools (15). Error codes guide with types and recovery advice, now with a 400 service_tier error for tiers a project doesn't allow (15). Dated model snapshots and a dated changelog, last entry 30 October (15).

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).

OpenAI API 98

Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Tool use with parallel calls and forced tool choice, 18 of 20 because GPT-6 Astra, GPT-6.1 Sol and GPT-6.1 Terra call tools only through the Responses API (18). Strict structured outputs (15). Automatic prompt caching with cached input at 0.1x on Astra and Luna ($0.01 per million on Luna) and 0.05x on GPT-6.1 Sol ($0.10) and GPT-6.1 Terra ($0.05), and a 12-hour cache lifetime in the Responses API since 20 October (15). 1.05M context on Astra, GPT-6.1 Sol, GPT-6.1 Terra and Luna (15). Batch at half price (10). Official SDKs in Python, JavaScript and more (10). Documented error codes with retry guidance (15).

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.

OpenAI API 87

Model reading. GPT-6 Rosalind, a life-sciences model for approved organisations, launched on 30 October, after GPT-6.1 Terra on 26 October and the Decisions API GA on 22 October (30). Published minimum notice of 6 months for GA models, 3 months for specialised variants, as little as 2 weeks for previews. On 14 October GPT-4o (gpt-4o-2024-08-06, gpt-4o-2024-11-20), GPT-4o mini, GPT-4.1 and GPT-4.1 mini were given a shutdown date of 14 January 2027, 92 days, with no safety or compliance reason stated. These are GA models, so the page's own 6-month minimum wasn't kept, as with gpt-5.4-cyber's 20 days in September. 8 of 12, the award for a minimum of two to three months in practice. Five shutdown dates in the 90 days to 1 November, the Assistants API on 26 August (announced in August 2025), gpt-5.2-chat-latest and gpt-5.3-chat-latest on 10 August (announced 8 May), gpt-3.5-turbo-instruct, babbage-002, davinci-002 and gpt-3.5-turbo-1106 on 28 September (announced 26 September 2025, missed in the 27 October count), gpt-5.4-cyber on 1 October and the legacy snapshots on 23 October (announced 22 April). Sora 2 and the Videos API (24 September) are scored on the openai-sora listing. Four of the five kept the published notice, so 4 of 8 as before, as amazon-bedrock has for four dates after full notice. Dated changelog several times a month and a help centre (15 of 15). openai-python had 215 open issues per the 1 October check, replies not sampled (5 of 10). Current official SDKs, openai 3.42.0 on PyPI (31 October) and npm openai 7.47.0 (15). Supported runtimes stated, Python 3.10 or later and Node 22 or later, the npm package's engines re-read on 1 November (10).

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).

OpenAI API 75

Status page at status.openai.com with component history (20). Two API-wide majors in the 90 days to 1 November, elevated errors across ChatGPT, Codex and the API (the Agents API included) on 29 September, with Chat Completions, Responses, Batch, Realtime, Embeddings, Images, Files and the other API components degraded from 17:52 to 23:14 UTC, and elevated error rates across API models on 17 September, with eleven components degraded from 20:20 to 21:20 UTC. Two majors give 5 of 30. The 25 July incident is now outside the window. The 29 September write-up, posted on 6 October, traces it to a feature rollout that exhausted shared network connection capacity. Every other incident since 3 August that lists an API component is partial or under an hour, so minor. Degraded performance across ChatGPT, Codex and the APIs on 21 October listed every API component but ran 22 minutes (02:26 to 02:48 UTC). Single-model incidents were GPT-5.6 Sol errors on 11 September (09:08 to 11:21 UTC), gpt-5.6-luna errors on 21 October (70 minutes) and GPT-6.1 Sol degraded performance on Batch, Chat Completions and Responses on 27 October (78 minutes). Partial incidents were latency for some Responses requests on the US endpoint on 31 August and 1 September, image generation where some requests failed on 8 September, latency for some Responses and Chat Completions requests on 30 September, large-PDF errors in Responses and an Agents API computer use incident on 6 October, Responses errors for users in India, Japan and Korea on 22 October and errors on some GPT Image 2 edit requests on 30 October (14:35 to 18:36 UTC). Image model incidents on 16 September and 14 October and file download errors on 19 October each ran about 40 minutes. FedRAMP errors for Astra and Sol on 21 October and authentication errors near Illinois on 20 October list no API component. Rate limits published per tier and model family with rate-limit headers (15). 429 guidance, re-read on 1 November, covers Retry-After as a minimum wait, backoff with jitter when it's missing, a gradual ramp, and splits slow_down (429) from server_is_overloaded (503) (15). Scale Tier carries a 99.9% uptime SLA, through sales, per the 1 October check, not re-read on 1 November (10). Responses and Chat Completions are GA (10). 20 + 5 + 15 + 15 + 10 + 10 = 75.

Amazon Bedrock 85

Closed service under the AWS Customer Agreement and Service Terms, with the SDKs and Smithy models under Apache-2.0 (18, as for S3). The data retention, abuse detection and security pages agree that Bedrock stores nothing by default, name the models that retain traffic and for how long, and say retained data isn't shared with providers. The retention page's default mode says AWS 'may retain the data for safety and abuse-prevention purposes' without a period, and a legacy provider_data_share mode is kept though sharing isn't supported (25 of 30). A written lifecycle policy with Legacy and EOL dates per model, but the legacy table's 34-day Legacy periods for five Claude models in some Regions contradict its own 6-month floor (15 of 20, down from 18). AWS's sub-processor list, per the 4 October S3 check, and per-Region, geographic and global inference options, with global routing able to process data in any commercial Region (18 of 20).

OpenAI API 90

Closed service with clear terms, SDKs Apache-2.0 (15). The data-controls guide, re-read on 1 November, gives retention periods per endpoint, the training default and the ZDR route. The DPA, effective 1 January 2026, commits to return or deletion of customer data on termination, 30 days to object to a new subprocessor, and SCCs with the UK Addendum for transfers, and nothing in it conflicts with the guide, per the 4 October check, not re-read on 1 November (30). Deprecations page with announcement and shutdown dates, and GET /v1/models returns each model's shutdown_date (20). The subprocessor list, updated 9 July 2026, names its third-party subprocessors and OpenAI affiliates with the processing countries for each and a form to get notice of new ones, and data residency regions are listed, per the 4 October check, not re-read on 1 November (20).