Revolut Says Self-Hosted GLM Cut Its AI Coding Costs 90%

Revolut moved 2,000 engineers off metered AI APIs onto self-hosted open models running on 56 Nvidia B200 GPUs, and says that cut its AI coding costs 90%. David Abitbol of Revolut gave the figures in a talk posted on the AI Engineer YouTube channel on 29 October, StartupHub.ai reported.

Abitbol said in the talk that self-hosting GLM brings the bank’s blended cost to 30 cents per million tokens, and that its invoice is one tenth of what it was spending on Anthropic and OpenAI; it now serves GLM-5.3. (Corrected 3 November: an earlier version named the model GLM 4.5, as StartupHub reported.) By our arithmetic, that puts the earlier cost near $3 per million tokens. For comparison, Anthropic lists Claude Sonnet 5.5 at $2 per million input tokens.

Business Pill · Lead Time Explained

A short explainer on lead time as a business concept. It teaches the general idea only and says nothing about this story.

The key insight: As we read the reports, Revolut and Decagon each kept frontier models for a narrow slice of work and moved routine, repeatable tasks to open models they run or tune themselves.

What Revolut Reported

StartupHub reports that the system “was demoed as a production coding assistant for Revolut’s engineers, not a lab test”. Its performance target was “sub 3 seconds to first token and about 200 tokens per second for interactive use”.

The results came in stages. StartupHub reports that “Gemini alone delivered roughly 10% faster lead time and 20% faster coding time in early 2025, while early Claude Code with strict budget caps showed little impact”.

After self-hosting GLM in April, StartupHub says, “Revolut saw about 60% improvement in coding time and 70% in lead time on a 2,000 engineer sample”. Time to first token fell “from over 10 seconds to 1.6 seconds” while requests per second rose “from 3 to over 7”.

Abitbol attributed the larger lead-time gain to work outside coding. StartupHub’s account: “like inbox triage and meetings, that engineers now automate when tokens are not rationed”.

Improvement Revolut's David Abitbol reported, % faster: Gemini alone in early 2025 (coding time 20%, lead time
Improvement Revolut’s David Abitbol reported, % faster: Gemini alone in early 2025 (coding time 20%, lead time 10%) vs self-hosted GLM from April (coding time about 60%, lead time 70%), on a 2,000-engineer sample, as reported by StartupHub.ai on 29 October 2026. Self-reported figures.

What Was Posted on X

Oliver Molander pointed to Revolut and Decagon in this 1 November post. Molander wrote: “I wouldn’t be surprised if 80–90% of routine enterprise AI workflows eventually run on open-weight and smaller models.” Revolut’s figures in this piece come from StartupHub.ai’s account of David Abitbol’s talk; Decagon’s split comes from Jesse Zhang’s own post.

Zain, who works on AI/ML at Together AI according to their X bio, posted a thumbnail of Revolut’s AI Engineer New York talk, titled “A Tenth of the Bill”, on 9 November. The slide in the image says 2,000 engineers are served with GLM 5.3 on 56 B200 GPUs “with Together”, and the post credits open models with a tenth of the coding-agent bill and products shipped 70% faster. Together AI appears in the talk as Revolut’s provider, so the post comes from a vendor employee.

What It Took

The savings came after tuning. StartupHub reports that the team “improved performance 88% for 76% more demand without adding GPUs by tuning context limits, quantization and parallelism”. It capped context at 256k, and NVFP4 on Blackwell “saves about 40% GPU footprint versus FP8”.

The 30-cent figure has limits. StartupHub notes that “the talk left cloud spend, staffing and observability costs outside the 30 cent token math”. Image tasks still go to Gemini or similar models, so the self-hosted path is text only.

By our arithmetic, 2,000 engineers on 56 GPUs is about 36 engineers per B200.

Decagon Says 90% Runs on Open Models

Revolut is not the only company to report this shift. Decagon co-founder Jesse Zhang wrote on X on 24 August that “Decagon’s workflows are now roughly split 90% open source, 10% closed source”. He added that “you do need a minimum amount of frontier tokens for new or novel projects”.

On The a16z Show, StartupHub reported, Zhang said frontier labs “do have small models, but you can’t really control them in the way that you want”. Co-founder Ashwin Sreenivas said of fine-tuned smaller models: “It is better at the task, it is cheaper, and it is faster.”

The Business Lever: Price per Token

Oliver Molander pointed to both companies in a post on X on 1 November. He wrote: “I wouldn’t be surprised if 80–90% of routine enterprise AI workflows eventually run on open-weight and smaller models.” He added that frontier labs do not lose, because “the market is splitting into two massive opportunities: frontier intelligence and specialized execution at scale”.

As we read the two cases, both moved routine, repeatable work off frontier APIs and kept frontier models for the rest: Revolut for images, Decagon for new or novel projects. Our token price table lists Claude Opus 5.5 at $4 and Claude Sonnet 5.5 at $2 per million input tokens. Revolut’s reported 30 cents is its own serving cost, which excludes cloud spend, staffing and observability, while those are list prices, so the two are not like for like.

The Structural Read

As we read StartupHub’s account, Revolut’s cost cut and its speed gain arrived together: time to first token fell from over 10 seconds to 1.6 seconds as it moved to self-hosted GLM.

As we read StartupHub’s account, the gain depended on tuning: Revolut reports 88% more performance for 76% more demand without adding GPUs.

As we read Zhang’s post, Decagon frames the split as a balance, not a replacement: it says you need a minimum amount of frontier tokens for new or novel projects.

Jesse Zhang, Decagon co-founder, on X, 24 August 2026

“Decagon’s workflows are now roughly split 90% open source, 10% closed source.”

Three Implications

FOR ANTHROPIC AND OPENAI StartupHub reports Revolut’s prior Anthropic and OpenAI invoice was ten times its self-hosted cost; Decagon says frontier tokens remain needed for new or novel projects.

FOR GPU CLOUDS AND NVIDIA Revolut reports running 2,000 engineers on 56 Nvidia B200s, about 36 engineers per GPU by our arithmetic.

FOR ENTERPRISE AI BUYERS StartupHub notes the 30-cent figure left cloud spend, staffing and observability costs outside the math, and that portability to smaller teams is unproven.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Open vs Closed Meta-Framework.

The framework says the winning pattern is to “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)”.

As we read it, Revolut and Decagon describe buyers treating routine model work as the abundant layer. The framework predates these reports and does not discuss either company.

Related coverage: AI Token Prices 2026 · Anthropic draws 2.1x OpenAI’s spend per token on OpenRouter · This Week in the AI Supercycle

Update (2 Nov): Westpac says AI halved a data migration’s cost

Westpac said an AI-accelerated migration of its legacy St.George data warehouse would cost around $35 million against a $70 million budget, iTnews reported on 2 November from the bank’s annual results. Chief executive Anthony Miller said the team used AI tools to speed up code conversion, testing and documentation: “This helped us complete the migration more than two years earlier at half the cost.” Westpac expects around $15 million a year in direct benefits and to start booking them from FY27, according to iTnews. It is a second enterprise data point after Revolut, set against Apollo economist Torsten Slok’s view that AI tokens are a cost centre outside tech.

Update (3 Nov): Revolut’s own talk names GLM-5.3 and calls the frontier “the exception”

The Register reported on 3 November, from Abitbol’s talk at the AI Engineer conference in New York, that Revolut cut its AI coding inference costs by roughly 90 percent in less than a year. We have now gone to the talk itself, posted by AI Engineer on 29 October. In it, Abitbol says Revolut started self-hosting open-source models in April 2026 with GLM-5.1, and that serving about 2,000 engineers takes 56 Nvidia B200s to host GLM-5.3, from China’s Z.ai. The Register gives the same sequence. An earlier version of this piece named GLM 4.5, following StartupHub; we have corrected it.

In the talk, Abitbol puts Revolut’s blended cost at 30 cents per million tokens and its invoice at one tenth of what it was spending on Anthropic and OpenAI. Before the switch, he said, some engineers were spending $1,000 or $10,000 a month with little return, so Claude Code launched with strict budget limits. “We were highly skeptical about the cost of AI,” he said. By our arithmetic, at the 30-cent blended rate, $1,000 a month would buy about 3.3 billion tokens.

Revolut still uses the frontier labs’ tools, as the exception. Abitbol said it still uses Codex and Claude Code and that “the latest models are fantastic”, but keeps them as the exception. As The Register quoted him: “Open source is now a priority at Revolut, and our economics are favorable. The billing is flat, we don’t have any bill shock at the end of the month, and the frontier is the exception.”

The gain came from tuning, not hardware. Between April and August, Abitbol said, requests per second rose from around three to over seven while time to first token fell from over 10 seconds to around 1.6, on the same 56 GPUs: an 88% improvement in performance for 76% more demand, in his figures. His levers were a context cap of 256,000 tokens (“long context is a scam”), FP4 quantization on Blackwell, tensor parallelism and a cache hit rate above 90% for coding tasks.

Revolut self-hosted GLM: time to first token fell from over 10 seconds to about 1.6 as requests per second rose from about 3 to over 7, same 56 B200 GPUs
Revolut’s serving metrics on the same 56 B200 GPUs, April vs August 2026, as David Abitbol stated them in his AI Engineer talk (spoken approximations). Source: Revolut talk, AI Engineer YouTube, 29 October 2026.

On compliance, he said Revolut takes the weights and runs them on US-based infrastructure rather than using inference endpoints in China, and keeps a human in the loop through a sandbox. GLM is text only, so image tasks are routed to other models such as Gemini.

Today’s other data points run the same way: LinkedIn swapped out OpenAI models for its own, and Menlo Ventures found open-source models at 19% of tokens, while Claude Code and Codex take 70% of $34 billion in AI coding spend.

What Is Not Established

The 3 November update draws on Revolut’s talk itself, from our machine transcription of the AI Engineer video, and on The Register; earlier sections rely on StartupHub’s account of the talk. The GLM versions (GLM-5.1, then GLM-5.3) are as Abitbol named them and The Register reported. The 3.3 billion tokens is our arithmetic and ignores the commercial tools’ own prices. The 30-cent cost excludes cloud spend, staffing and observability, by StartupHub’s account. Decagon’s 90% is the founders’ description of its workflows, without a stated measure (tokens, requests or spend). Neither company’s figures have been independently audited. The $3 prior cost and 36 engineers per GPU are our arithmetic.

This Week In The AI Supercycle

Anthropic’s Reality Check, The Capital Squeeze, & The New World Order

This week’s issue: for two years the AI build-out was priced as a growth story. This week it started being priced as a credit story.

Read the full issue →

Business Engineer Framework

The Open vs Closed Meta-Framework

A Business Engineer framework on which layers of the AI stack stay proprietary and which get commoditized.

Read the Map of AI →

The Bottom Line

Revolut says self-hosting GLM on 56 Nvidia B200s cut its AI coding costs 90%, to about 30 cents per million tokens for 2,000 engineers, according to StartupHub’s account of a talk posted on 29 October. Decagon’s co-founder wrote in August that its workflows run roughly 90% on open-source models.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Revolut’s figures come from StartupHub.ai’s 29 October account of David Abitbol’s talk on the AI Engineer YouTube channel; we did not watch the talk and did not contact Revolut. Decagon’s split comes from Jesse Zhang’s 24 August post on X and StartupHub’s report of the founders’ a16z interview. Token list prices are from our own price table. The prior-cost and per-GPU figures are our arithmetic.

Sources: StartupHub.ai, Revolut says self-hosted AI cut coding costs 90%, 29 October 2026 · AI Engineer on YouTube, How Revolut Cut AI Coding Costs 90% With Self-Hosted Open Models – David Abitbol, Revolut · Jesse Zhang on X, 24 August 2026 · StartupHub.ai, Why Decagon Runs 90 Percent of Its Enterprise AI on Open Source, 31 July 2026 · Oliver Molander on X, 1 November 2026 · FourWeekMBA, AI Token Prices 2026

Scroll to Top
FourWeekMBA