{"data":{"items":[{"id":"98b9dcca-623c-40ef-bc0c-5e919103c425","excerpt":" — Revolut made self-hosted open-weight GLM models the default coding AI for ~2,000 engineers, on 56 Nvidia B200s. Its bill is now about a tenth of what it paid Anthropic and OpenAI.\n\nWorth copying: on the same GPUs, tuning cut time to first token from 14.3s to 1.7s while demand rose 76%.","url":"https://bsky.app/profile/karanluthra.bsky.social/post/3mzi6gvhbyt2f","role":"pricing","weight":0.596929,"occurredAt":"2026-11-04T10:47:06.879Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.48139435,"matchedPatterns":["free_tier","product:anthropic"],"statement":"Revolut made self-hosted open-weight GLM models the default coding AI for ~2,000 engineers, on 56 Nvidia B200s.","title":null,"body":"Revolut made self-hosted open-weight GLM models the default coding AI for ~2,000 engineers, on 56 Nvidia B200s. Its bill is now about a tenth of what it paid Anthropic and OpenAI.\n\nWorth copying: on the same GPUs, tuning cut time to first token from 14.3s to 1.7s while demand rose 76%.","authorHash":"9a9b615b44970ce63edb5737afa61db0","rating":null,"offTopic":true},{"id":"903f9e0a-b2db-47aa-bc91-8ad8f76ed057","excerpt":" — Revolut moved ~2,000 engineers off Claude Code and Codex to self-hosted GLM on 56 B200s. Bill ~10x lower.\n\nThe interesting part: their productivity gains showed up once token limits were lifted. Agents work better when tokens aren’t rationed.","url":"https://bsky.app/profile/rtk-ai.app/post/3mzia2dam2s2c","role":"pricing","weight":0.5671218,"occurredAt":"2026-11-04T11:15:53.343Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.45735627,"matchedPatterns":["free_tier","product:claude"],"statement":"Revolut moved ~2,000 engineers off Claude Code and Codex to self-hosted GLM on 56 B200s.","title":null,"body":"Revolut moved ~2,000 engineers off Claude Code and Codex to self-hosted GLM on 56 B200s. Bill ~10x lower.\n\nThe interesting part: their productivity gains showed up once token limits were lifted. Agents work better when tokens aren’t rationed.","authorHash":"bdd0a0cef2f847646f63398aef3c5a0c","rating":null,"offTopic":true},{"id":"7b235126-7625-4512-b522-6d9dc46fbe5d","excerpt":" — revolut moved ~2,000 engineers to self-hosted glm 5.3 on 56 b200s 🏦\n\n~$0.30 per 1m tokens, about a tenth of the old anthropic/openai bill. claude code + codex are now the exception\n\nunlimited cheap tokens > rationed smart ones 👀","url":"https://bsky.app/profile/daniakash.com/post/3mzet2a5y452c","role":"pricing","weight":0.49651667,"occurredAt":"2026-11-03T02:45:10.209Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.40041667,"matchedPatterns":["free_tier","product:anthropic"],"statement":"revolut moved ~2,000 engineers to self-hosted glm 5.3 on 56 b200s 🏦 ~$0.30 per 1m tokens, about a tenth of the old anthropic/openai bill.","title":null,"body":"revolut moved ~2,000 engineers to self-hosted glm 5.3 on 56 b200s 🏦\n\n~$0.30 per 1m tokens, about a tenth of the old anthropic/openai bill. claude code + codex are now the exception\n\nunlimited cheap tokens > rationed smart ones 👀","authorHash":"979f8ef6af8781487c4c74570347d1e3","rating":null,"offTopic":true},{"id":"1b234f05-a282-40ed-adfa-ef655b4e1d48","excerpt":" — ‼️ Revolut says it cut AI coding costs about 10x by making a self-hosted Chinese model its default in place of Claude Code.\n\nIt runs Z ai's GLM on its own Nvidia GPUs for about \\$0.30 per million tokens.","url":"https://bsky.app/profile/intcyberdigest.bsky.social/post/3mzjraa54qg2p","role":"pricing","weight":0.49263993,"occurredAt":"2026-11-05T01:56:04.152Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.39729026,"matchedPatterns":["free_tier","product:claude"],"statement":"‼️ Revolut says it cut AI coding costs about 10x by making a self-hosted Chinese model its default in place of Claude Code.","title":null,"body":"‼️ Revolut says it cut AI coding costs about 10x by making a self-hosted Chinese model its default in place of Claude Code.\n\nIt runs Z ai's GLM on its own Nvidia GPUs for about \\$0.30 per million tokens.","authorHash":"488880b7a27b06878706b28c0ecaeb79","rating":null,"offTopic":true},{"id":"d344f46e-672e-4313-bdf2-7e135abc2a35","excerpt":" — Revolut now does AI coding entirely with self-hosted, open weight models (like GLM-5.3). Cost savings look impressive but the article doesn’t factor in how much they had to pay for their Nvidia GPUs. Hmm […]","url":"https://bsky.app/profile/kumarvibe.mastodon.social.ap.brid.gy/post/3mzi6plrrzdr2","role":"pricing","weight":0.47373167,"occurredAt":"2026-11-04T10:51:49.000Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0.6666667,"confidence":0.38204166,"matchedPatterns":["free_tier","solution:self-hosted"],"statement":"Revolut now does AI coding entirely with self-hosted, open weight models (like GLM-5.3).","title":null,"body":"Revolut now does AI coding entirely with self-hosted, open weight models (like GLM-5.3). Cost savings look impressive but the article doesn’t factor in how much they had to pay for their Nvidia GPUs. Hmm […]","authorHash":"cb7d6f2e2eaee274f3e9e2aeeb52efbc","rating":null,"offTopic":false},{"id":"08279c43-e9a8-473c-8c31-73298a8596fd","excerpt":" — Revolut's own figures: self-hosting Chinese GLM models cut its AI coding inference bill by about 90%. It runs them on 56 GPUs, so smaller firms shouldn't assume the same saving.\n\nhttps://resultsense.com/r/2026-11-04-revolut-self-hosted-glm-ai-coding-costs\n\n#EnterpriseAI","url":"https://bsky.app/profile/resultsense.bsky.social/post/3mzl3eze73y2j","role":"pricing","weight":0.429839,"occurredAt":"2026-11-05T14:30:22.181Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.34664434,"matchedPatterns":["free_tier","solution:hosting"],"statement":"Revolut's own figures: self-hosting Chinese GLM models cut its AI coding inference bill by about 90%.","title":null,"body":"Revolut's own figures: self-hosting Chinese GLM models cut its AI coding inference bill by about 90%. It runs them on 56 GPUs, so smaller firms shouldn't assume the same saving.\n\nhttps://resultsense.com/r/2026-11-04-revolut-self-hosted-glm-ai-coding-costs\n\n#EnterpriseAI","authorHash":"510d830d7d6009ba40548f819e20b6ee","rating":null,"offTopic":true},{"id":"36912284-19ee-4514-99c1-1393a4ecc557","excerpt":" — Revolut AI coding costs cut 90% with self-hosted GLM on 56 B200s\n\nhttps://www.startuphub.ai/ai-news/artificial-intelligence/2026/revolut-says-self-hosted-ai-cut-coding-costs-90","url":"https://bsky.app/profile/startuphub.bsky.social/post/3myyyi4vh4u22","role":"pricing","weight":0.31857666,"occurredAt":"2026-10-29T09:50:30.185Z","sourceKey":"bluesky","sourceName":"Bluesky","credibility":0.5,"venue":"selfhost-homelab","intent":"pricing_complaint","painScore":0.24,"sentiment":0,"confidence":0.25691667,"matchedPatterns":["free_tier","solution:self-hosted"],"statement":"Revolut AI coding costs cut 90% with self-hosted GLM on 56 B200s","title":null,"body":"Revolut AI coding costs cut 90% with self-hosted GLM on 56 B200s\n\nhttps://www.startuphub.ai/ai-news/artificial-intelligence/2026/revolut-says-self-hosted-ai-cut-coding-costs-90","authorHash":"ddd7574b11a2f75254be343b345cec88","rating":null,"offTopic":true}],"breakdown":[{"sourceKey":"bluesky","sourceName":"Bluesky","count":7}],"total":7}}