Inkling 2 Mini on a 4090: catches citations near-Haiku level, 2-3x slower than Qwen
strickvl · x · 2026-11-21
A blogger overnight-tested Thinky Machines' new Inkling 2 Mini on a historian-in-a-box citation-checking pipeline running on a single RTX 4090:
- Accuracy: missed only 6 of 42 wrong citations — near Claude Haiku level (2), well ahead of Qwen3.6 35B (19) and Gemma 4 26B (14). Other tasks (profile updates, fact filing) were on par with the comparison models.
- Downside: reasoning can't be fully disabled, making the whole pipeline 2-3x slower than Qwen (308s vs 90s).
- Verdict: the author is switching the citation checker's default to Inkling 2 Mini — accuracy beats speed here — and is excited about what capable local small models unlock.
More from Models
- Google reportedly changing privacy and AI training opt-in controls — blackice193 · 2026-11-21
- Tao announces AI-assisted proofs: Euler/Navier-Stokes blowup and Generalized Riemann Hypothesis — burny_tech · 2026-11-21
- kale-tts-nano pitch movement holds to the end: 2.83 st early vs 2.87 st late in sentences — NERDDISCO · 2026-11-21
- 7.5M TTS model holds prosody on long sentences: kale-tts-nano f0 std 2.97 st matches 45.7M sibling — NERDDISCO · 2026-11-21
- OVH's Olala Mini 7A1B: 360 tok/s decode on one RTX 5090, 96K context, v2.0 to exceed 100B — JiliJeanlouis · 2026-11-21
- 6-round eval shows GPT-6.2-sol matching only 11/23 judge outcomes vs 6.1's 21/23 — 0xkarasy · 2026-11-21