DoubleChecks

What was claimed

Inkling-Small is the strongest open model for its size, achieves comparable performance to the larger Inkling at a quarter of its size, and runs locally on 128GB RAM with image, audio, and 1M context support

Our verdict

Needs caution

Artificial Analysis reports that no open-weights model at Inkling-Small’s size or smaller scores higher on its Intelligence Index, and it explicitly compares within that framework. This supports "strongest" on that specific benchmark, but calling it the strongest open model for its size as a general fact over all tasks and evaluations is too broad and not independently corroborated beyond that single index. Available sources describe the model’s parameter counts, performance, and open-weights status but do not state specific RAM requirements for local inference. Without official deployment specs or hardware guidance mentioning 128GB RAM, this claim cannot be confirmed or refuted.

All 3 AI systems agree20 sources citedChecked Jul 31, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

Inkling-Small is the strongest open model for its size

Misleading82%
2 of 3 AIs agree·ChatGPT: Can’t verify

Inkling-Small runs locally on 128GB RAM

Can’t verify63%
2 of 3 AIs agree·Claude: Misleading

Inkling-Small has image, audio, and 1M context support

Verified91%
All 5 AIs agree

Inkling-Small achieves comparable performance to the larger Inkling at a quarter of its size

Verified93%
2 of 3 AIs agree·ChatGPT: Misleading

Verified against 20 sources

thinkingmachines.aiartificialanalysis.aigigazine.netartificialanalysis.aiventurebeat.comx.comecorpit.comnote.comiotdigitaltwinplm.comventurebeat.comxenospectrum.comawesomeagents.aiartificialanalysis.aigraphify.netreddit.comsebastianraschka.comreddit.comyoutube.comyoutube.comhuggingface.co

Detailed Analysis

Most elements of the claim are broadly supported by current sources, but parts are overstated or use imprecise language. The statement that it is the strongest open model at its size is based on a single benchmarking source and should be treated as contextual, not absolute. The rest about relative size, performance, and multimodal/1M context support is accurate, while the RAM requirement is not specified in sources and is therefore unverified.

Why this verdict

  • Most elements of the claim are broadly supported by current sources, but parts are overstated or use imprecise language.
  • The statement that it is the strongest open model at its size is based on a single benchmarking source and should be treated as contextual, not absolute.
  • The rest about relative size, performance, and multimodal/1M context support is accurate, while the RAM requirement is not specified in sources and is therefore unverified.

Claims checked

Inkling-Small is the strongest open model for its size

Misleading78%
Artificial Analysis reports that no open-weights model at Inkling-Small’s size or smaller scores higher on its Intelligence Index, and it explicitly compares within that framework.[4] This supports "strongest" on that specific benchmark, but calling it the strongest open model for its size as a general fact over all tasks and evaluations is too broad and not independently corroborated beyond that single index.[4] [2] [13]

This claim may mislead readers without added context.

Inkling-Small runs locally on 128GB RAM

Can’t verify55%
Available sources describe the model’s parameter counts, performance, and open-weights status but do not state specific RAM requirements for local inference.[1] [4] [5] [14] [19] Without official deployment specs or hardware guidance mentioning 128GB RAM, this claim cannot be confirmed or refuted.

Inkling-Small achieves comparable performance to the larger Inkling at a quarter of its size

Verified90%
Thinking Machines and multiple analyses state that Inkling-Small matches or exceeds Inkling on many benchmarks, while lagging on some agentic and factual tasks.[1] [4] [5] [9] [11] [12] VentureBeat and Gigazine explicitly describe it as nearing or equivalent to Inkling’s performance at about one-quarter of its size.[3] [5]

What other AI think

The claim contains mostly accurate information about Inkling-Small's capabilities and performance, but makes an absolute claim about being "the strongest open model for its size" that is not fully supported by evidence. Additionally, the 128GB RAM specification is only accurate for certain quantized versions, not the full model.

Why this verdict

  • The claim contains mostly accurate information about Inkling-Small's capabilities and performance, but makes an absolute claim about being "the strongest open model for its size" that is not fully supported by evidence.
  • Additionally, the 128GB RAM specification is only accurate for certain quantized versions, not the full model.

Claims checked

Runs locally on 128GB RAM

Misleading80%
Inkling-Small dynamic 3-bit runs on a 128GB RAM device (NVIDIA DGX Spark or Mac) and the smaller 4-bit runs on 132GB RAM. This is only true for quantized versions, not the full BF16 model, which requires significantly more resources.

This claim may mislead readers without added context.

Inkling-Small is the strongest open model for its size

Misleading85%
No open weights model at its size or smaller scores higher on the Intelligence Index, but this is based on one specific benchmark. DeepSeek V4 Flash (max), at a similar size of 284B total (13B active) also scores 40, matching Inkling-Small's score, so the claim of being 'the strongest' is overstated.

This claim may mislead readers without added context.

Inkling-Small achieves comparable performance to the larger Inkling at a quarter of its size

Verified95%
Inkling-Small achieves comparable performance to Inkling at a quarter of its size, and Inkling-Small matches or exceeds Inkling on reasoning and agentic tasks.
Some parts of the statement match public sources about Inkling and its smaller sibling, but other parts are subjective or imprecise and one hardware claim is not verifiable. Numerical size-comparison and local-run claims are misleading or unproven.

Why this verdict

  • Some parts of the statement match public sources about Inkling and its smaller sibling, but other parts are subjective or imprecise and one hardware claim is not verifiable.
  • Numerical size-comparison and local-run claims are misleading or unproven.

Claims checked

Inkling-Small achieves comparable performance to the larger Inkling at a quarter of its size

Misleading80%
Benchmarks reported by the project show Inkling-Small matches or exceeds the larger model on some tasks, but its total/active parameter counts (≈276B total / 12B active vs ≈975B total / 41B active) correspond to ~28–30% of active parameters, not exactly one quarter, so the "quarter of its size" phrasing is imprecise.

This claim may mislead readers without added context.

Inkling-Small is the strongest open model for its size

Can’t verify60%
"Strongest" is subjective and requires a defined benchmark suite and comparisons; public previews and community posts praise Inkling-Small but do not provide a definitive, reproducible ranking.

Inkling-Small runs locally on 128GB RAM

Can’t verify70%
Community reports discuss attempts and optimizations, but official weights and precise memory requirements (with quantization/offloading) are not fully published, so it cannot be confirmed that a typical 128 GB machine can run the full model reliably.

Share this result

Try DoubleChecks