Inkling vs Qwen3 235B A22B (2507)

Thinking Machines Lab · US  |  Alibaba · China · Updated June 2026

Quick verdict

Pick Inkling for the first model release from mira murati's thinking machines lab (ex-openai cto) - one of the most closely-watched new labs in ai or a 975b-parameter moe (41b active) with native text, image, and audio reasoning in one model. Pick Qwen3 235B A22B (2507) for deep world knowledge from 235b total parameters (83.0 mmlu-pro, 93.1 mmlu-redux) or exceptional multilingual and alignment results (79.2 arena-hard v2, 85.2 writingbench).

Inkling (Thinking Machines Lab, US) and Qwen3 235B A22B (2507) (Alibaba, China) line up two different AI ecosystems against each other — a comparison that is as much about cost philosophy and openness as raw capability. Inkling is mira Murati's Thinking Machines Lab debuts with Inkling - a 975B open-weight MoE with native multimodal reasoning and a dialable thinking-effort control. Qwen3 235B A22B (2507) is an older 235B text-only open mixture-of-experts with broad knowledge and strong writing — but no vision, no thinking mode, and weak coding. Their biggest split is context window, and the breakdown below shows exactly how that plays out for your workload.

Key differences at a glance

Side-by-side specs

SpecInklingQwen3 235B A22B (2507)
ProviderThinking Machines Lab (US) Alibaba (China)
ReleasedJuly 15, 2026 July 21, 2025
Context window1M (~1,500 pages) 256K (~393 pages)
Price (in/out)Open weight (self-host / free) Open weight (self-host / free)
Open weight?Yes — self-hostable Yes — self-hostable
Modalitiestext, image, audio, code text, code
SWE-Bench VerifiedNot published Not published
MRCR v2 @ 1MNot published Not published

Who wins what

The first model release from Mira Murati's Thinking Machines Lab (ex-OpenAI CTO) - one of the most closely-watched new labs in AI

Inkling

Qwen3 235B A22B (2507) is comparatively weak here — text-only with no vision, and the absence of a thinking mode caps its hardest reasoning

A 975B-parameter MoE (41B active) with native text, image, and audio reasoning in one model

Inkling

Mira Murati's Thinking Machines Lab debuts with Inkling - a 975B open-weight MoE with native multimodal reasoning and a dialable thinking-effort control — and it carries the larger 1M context.

A dialable 'thinking effort' knob (0.2-0.99) that lets you trade speed for depth of reasoning per-request

Inkling

Mira Murati's Thinking Machines Lab debuts with Inkling - a 975B open-weight MoE with native multimodal reasoning and a dialable thinking-effort control — and it is the newer of the two.

Deep world knowledge from 235B total parameters (83.0 MMLU-Pro, 93.1 MMLU-Redux)

Qwen3 235B A22B (2507)

Qwen3 235B A22B (2507) lists deep world knowledge from 235B total parameters (83.0 MMLU-Pro, 93.1 MMLU-Redux) among its strengths; Inkling does not.

Exceptional multilingual and alignment results (79.2 Arena-Hard v2, 85.2 WritingBench)

Qwen3 235B A22B (2507)

Qwen3 235B A22B (2507) lists exceptional multilingual and alignment results (79.2 Arena-Hard v2, 85.2 WritingBench) among its strengths; Inkling does not.

Outstanding structured logic — 95.0 on ZebraLogic

Qwen3 235B A22B (2507)

Qwen3 235B A22B (2507) lists outstanding structured logic — 95.0 on ZebraLogic among its strengths; Inkling does not.

Largest single-prompt input

Inkling

Its 1M window is about 3.8× larger than Qwen3 235B A22B (2507)'s 256K, fitting roughly 1,500 pages in one prompt.

Which should you pick?

Someone analysing very long documents or codebases

Inkling

Larger 1M window fits more in one prompt.

Anyone whose priority is the first model release from mira murati's thinking machines lab (ex-openai cto) - one of the most closely-watched new labs in ai

Inkling

It is specifically built for that.

Anyone whose priority is deep world knowledge from 235b total parameters (83.0 mmlu-pro, 93.1 mmlu-redux)

Qwen3 235B A22B (2507)

That is its strongest area.

An enterprise with regional data-residency rules

Inkling or Qwen3 235B A22B (2507)

Origin (US vs China) affects where data is processed and which compliance regime applies — check the provider's terms for your region.

Inkling: where it fits

Mira Murati's Thinking Machines Lab debuts with Inkling - a 975B open-weight MoE with native multimodal reasoning and a dialable thinking-effort control. Released July 15, 2026 by Thinking Machines Lab, it is built for the first model release from Mira Murati's Thinking Machines Lab (ex-OpenAI CTO) - one of the most closely-watched new labs in AI, a 975B-parameter MoE (41B active) with native text, image, and audio reasoning in one model, a dialable 'thinking effort' knob (0.2-0.99) that lets you trade speed for depth of reasoning per-request, and fully open weights (Apache 2.0) at frontier scale - unusual for a model this large and this new.

Its trade-offs are real: a brand-new lab's first release - no multi-generation track record yet, no official hosted API price - available via third-party hosts only at launch, and independent third-party benchmark verification is still limited given how recently it shipped. As an open-weight model, its running cost is your own hardware rather than a per-token fee.

Qwen3 235B A22B (2507): where it fits

An older 235B text-only open mixture-of-experts with broad knowledge and strong writing — but no vision, no thinking mode, and weak coding. Released July 21, 2025 by Alibaba, it is built for deep world knowledge from 235B total parameters (83.0 MMLU-Pro, 93.1 MMLU-Redux), exceptional multilingual and alignment results (79.2 Arena-Hard v2, 85.2 WritingBench), outstanding structured logic — 95.0 on ZebraLogic, and no thinking mode, which makes latency and token spend entirely predictable.

Its trade-offs: nearly a year old and superseded — Artificial Analysis now steers users to Qwen3.5-397B instead, text-only with no vision, and the absence of a thinking mode caps its hardest reasoning, coding is weak by 2026 standards, and it publishes no SWE-Bench score to compare on, and its 235B weights need roughly 438GB in BF16, far beyond consumer hardware. As an open-weight model, its running cost is your own hardware rather than a per-token fee.

The bottom line for this matchup

This is less "which is smarter" and more "which ecosystem fits." Inkling (US) and Qwen3 235B A22B (2507) (China) differ on pricing philosophy, data-residency, and tooling as much as on raw scores. The pragmatic move is to run one real task through both and judge the outputs against your own constraints — including where your data is allowed to be processed.

Want both Inkling and Qwen3 235B A22B (2507) without two subscriptions? LumiChats gives you these plus 40+ models under one ₹69/day pass (about $1/day) — draft with one, cross-check with the other.

See pricing

Frequently asked questions

Is Inkling or Qwen3 235B A22B (2507) better for coding?

Public SWE-Bench figures are not available for either model, so the honest test is your own repository — run an identical real bug through both. By design, Inkling leans toward the first model release from mira murati's thinking machines lab (ex-openai cto) - one of the most closely-watched new labs in ai while Qwen3 235B A22B (2507) leans toward deep world knowledge from 235b total parameters (83.0 mmlu-pro, 93.1 mmlu-redux), and that positioning usually predicts which feels better on your codebase.

Which is cheaper, Inkling or Qwen3 235B A22B (2507)?

They are priced almost identically, so cost will not decide between them.

Which has the bigger context window?

Inkling — 1M vs 256K, about 3.8× larger. Useful only if the model actually reasons over the full window, which not all do.

Can I use both Inkling and Qwen3 235B A22B (2507) together?

Yes — a multi-model platform like LumiChats gives you Inkling, Qwen3 235B A22B (2507) and 40+ others under one ₹69/day pass (about $1/day), so you can draft with one and cross-check with the other instead of buying two subscriptions.

Which is newer, Inkling or Qwen3 235B A22B (2507)?

Inkling — released July 15, 2026, about 12 months after Qwen3 235B A22B (2507).

Related comparisons

Specifications and benchmarks reflect publicly reported figures as of June 2026 and may change as providers release updates. Always verify on your own workload.