Inkling-2-Small weights released
Comments on “Inkling-2-Small weights released”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
Thinking Machines released open weights for Inkling 2 Small, a multimodal mixture-of-experts model with 594B total and 17B active parameters, a 1M-token context window and native NVFP4 precision. The NVFP4 checkpoint needs about 500GB of VRAM, for example two B300s or four H200s. The model card reports leading results on BrowseComp (90.2%) and several agentic benchmarks, and it is mixed elsewhere. All figures are the developer's own and none is independently verified. Support is already in vLLM, SGLang and several hosted providers.
Why it ranked
No. 9 of 137 with a score of 0.70. The panel rated it highest on new information (0.90) and practical relevance (0.82), lowest on significance (0.70). Placed No. 9, No. 17 by score alone: the placement terms moved other stories past it.
- AI judgeJudge 2claude-haiku-5-5
- Major open-weights multimodal LLM release with extensive benchmark tables and deployment details; substantial practical relevance for builders, though benchmarks are vendor-reported and largely unverified.
- AI judgeJudge 3gpt-5.6-luna
- A substantial open-weight multimodal model release with architecture, deployment requirements, extensive benchmark results, and safety documentation. It is highly actionable for AI builders, though some comparative claims rely on vendor-reported evaluations and promotional language.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.70total under rubric v2.3
- Significance0.70
- Reader interest0.79
- Audience breadth0.73
- Practical relevance0.82
- Substantiated0.77
- Technical depth0.73
- New information0.90
- Counts against: Promotional0.21
- Counts against: Overclaim0.31
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.72 | 0.60 | 0.78 | 0.70 | 0.18 | +0.24 | +0.168 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.63 | 0.80 | 0.93 | 0.79 | 0.30 | +0.26 | +0.205 |
| Audience breadthHow many people who build technology does it concern? | 0.70 | 0.72 | 0.78 | 0.73 | 0.08 | +0.13 | +0.095 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.83 | 0.80 | 0.84 | 0.82 | 0.04 | +0.22 | +0.181 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.93 | 0.55 | 0.82 | 0.77 | 0.38 | +0.05 | +0.038 |
| Technical depthHow technically substantive is it? | 0.84 | 0.60 | 0.76 | 0.73 | 0.24 | +0.05 | +0.037 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.91 | 0.85 | 0.94 | 0.90 | 0.09 | +0.05 | +0.045 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.10 | 0.35 | 0.18 | 0.21 | 0.25 | −0.19 | −0.040 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.16 | 0.40 | 0.36 | 0.31 | 0.24 | −0.11 | −0.034 |
| Sum of contributions | +0.696 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.696 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-13/9/"><img src="https://rubric.hrih.si/badge/2026-11-13/9.svg" alt="Ranked #9 on Rubric, 13 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 8 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenNumber
Inkling 2 Small has 594B total parameters, of which 17B are active.
The article gives no link or citation for it.
The article says “594B total, 17B active”
The source says No source to compare with.
No source givenNumber
Inkling 2 Small supports a context window of up to 1M tokens.
The article gives no link or citation for it.
The article says “It supports a context window of up to 1M tokens”
The source says No source to compare with.
No source givenNumber
Inkling 2 Small is a 48-layer decoder-only sparse Mixture-of-Experts transformer that routes each token to 8 of 384 experts.
The article gives no link or citation for it.
The article says “A 48-layer decoder-only transformer with a sparse Mixture-of-Experts (MoE) feed-forward backbone: each token is routed to 8 of 384 experts”
The source says No source to compare with.
No source givenNumber
The main NVFP4 checkpoint of Inkling 2 Small requires at least 500GB of aggregated VRAM, e.g. 2x B300 or 4x H200 GPUs.
The article gives no link or citation for it.
The article says “The main NVFP4 checkpoint requires a GPU cluster with at least 500GB of aggregated VRAM”
The source says No source to compare with.
No source givenCausal conclusion
The model was pretrained, midtrained and post-trained in NVFP4, making it the native precision.
The article gives no link or citation for it.
The article says “The model was pretrained, midtrained and post-trained in NVFP4, so this is the native precision for the model”
The source says No source to compare with.
No source givenNumber
The MXFP8 checkpoint gains only 1–2 percentage points over the NVFP4 checkpoint across evals.
The article gives no link or citation for it.
The article says “The MXFP8 has a very marginal 1–2 percentage points gain over the NVFP4 checkpoint over evals”
The source says No source to compare with.
No source givenBenchmark result
Inkling 2 Small scores 90.2% on BrowseComp (with compaction), above the other listed models except none higher than 89.4%.
The article gives no link or citation for it.
The article says “BrowseComp Inkling 2 Small with compaction 90.2%”
The source says No source to compare with.
No source givenGeneralization
Safety evaluations concluded Inkling 2 Small did not present risk of material uplift beyond what is already available in the open-weight ecosystem.
The article gives no link or citation for it.
The article says “we concluded that Inkling 2 Small did not present risk of material uplift beyond what's already available in the open-weight ecosystem”
The source says No source to compare with.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 15 Nov 2026, 20:42 UTC. It played no part in the ranking.