VQV Signal

PRODUCT UPDATE WATCH TECHNICAL

Exo Labs reports Inkling 2 Mini at 210 tokens/sec on M5 Ultra Mac Studio

According to a post on X by @exolabs, Exo Labs says Inkling 2 Mini reached 210 tokens per second in decoding on a 96GB M5 Ultra Mac Studio and 125 tokens per second on a 48GB M5 Max MacBook Pro. The company also says the model built a weather app 1.8x faster and 11x cheaper than Claude Sonnet 5.5. Separately, NDIF, Thinking Machines and Goodfire plan to launch a Looped Lens Challenge on Monday, aimed at surfacing what the looped model computes but never writes.

Source: Scobleizer News Pulse (X) · x.com Found via Robert Scoble’s Scobleizer News Pulse · a post on X, not independently verified Published 2026-11-21T02:45:00+00:00 Detected 2026-11-21T10:00:32+00:00
View original source

According to a post on X by @exolabs, Exo Labs says Inkling 2 Mini reached 210 tokens per second in decoding on a 96GB M5 Ultra Mac Studio and 125 tokens per second on a 48GB M5 Max MacBook Pro. The company also says the model built a weather app 1.8x faster and 11x cheaper than Claude Sonnet 5.5. Separately, NDIF, Thinking Machines and Goodfire plan to launch a Looped Lens Challenge on Monday, aimed at surfacing what the looped model computes but never writes.

AI-assisted summary based on the listed source.

Exo Labs says it reached 210 tokens/sec decoding on a 96GB M5 Ultra Mac Studio and 125 tokens/sec on a 48GB M5 Max MacBook Pro, and that the model built a weather app 1.8x faster and 11x cheaper than Claude Sonnet 5.5. NDIF, Thinking Machines and Goodfire launch a Looped Lens Challenge on Monday to surface what...

For people weighing local AI on Mac hardware, these speeds are company-reported claims and should be checked against independent tests before being relied on. The interpretability challenge targets a research question about what a looped model computes internally without showing it in its output. The source text is cut off, so the full reasoning behind the claims is not visible here; watch the challenge launch and any outside benchmarks.

Developers may have a new tool, API, or workflow change worth testing against their own stack.

Signal Strength 88% Technical label WATCH Public Interest 46 Category PRODUCT UPDATE Reader Depth TECHNICAL

Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.

Public Interest components
Recognizable Entity Score 51 Practical Impact Score 18 Novelty Interest Score 94 Consequence Score 34 Curiosity Score 0 Shareability Score 59

VQV surfaced this signal because it is recent, relevant to Consumer AI, connected to Scobleizer News Pulse (X).