Bloomberg: Gemini 4 shines in benchmarks but struggles in real-world coding, insiders say

Bloomberg reports that Google is preparing to launch its next-generation flagship model Gemini 4. While the model performs impressively on industry-standard benchmarks, anonymous insiders with direct access to it say employees remain skeptical in real-world use, with the model struggling in certain "key areas" such as coding—falling short of what its benchmark scores suggest—and internal opinions are divided on its true performance.

Confirmed

• Bloomberg published the report on October 1, citing Google insiders with direct access to Gemini 4 who requested anonymity because they were discussing internal matters
• The report says Gemini 4 performs strongly on industry benchmarks but underdelivers when put to work on real tasks such as coding
• Alphabet is preparing for the official launch of Gemini 4

Unconfirmed

• The information comes from an internal leak and has not been officially confirmed by Google
• Gemini 4's real-world performance after its consumer release remains to be seen; @Neurogence believes the situation may still change by the time of the official launch

Why it matters

• Coding capability is a key battleground in the current frontier-model race; if Gemini 4 truly lags behind on coding tasks, it could affect Google's competitive pace and product rollout against rivals
• The gap between benchmark scores and real-world experience has once again sparked industry debate over how much benchmarks should be trusted

2026-10-01 ~ 2026-10-01 · 5 related posts

• [source] Bloomberg: Gemini 4 aces benchmarks but struggles at real coding tasks internally — firstadopter · 2026-10-01
• [source] Gemini 4 aces benchmarks but struggles with real work, Google employees say — Neurogence · 2026-10-01
• Gemini 4 reportedly struggling on key areas like coding, per Bloomberg — ChrisGPT · 2026-10-01

2 near-duplicate retellings: firstadopter · ChrisGPT