Skip to content
TrustList
TQ
AI Model

ThinkingCap-Qwen3.8-27B

BottleCap AI's Qwen3.8-27B fine-tune that uses about a third fewer thinking tokens, licensed for small businesses

About ThinkingCap-Qwen3.8-27B

Reasoning models spend many tokens thinking before they answer, and ThinkingCap-Qwen3.8-27B is BottleCap AI's attempt to trim that. It is a fine-tune of Alibaba's Qwen3.8-27B. BottleCap's blog post is dated 22 September 2026, the weights reached Hugging Face on 23 September, and MarkTechPost reported the release on 23 and 24 September.

What stays the same

Like the model it is built on, it accepts text and images and returns text. Its served context is 262,144 tokens and it has 27.8 billion parameters. Nothing about the interface changes; what changes is how long it reasons.

Fewer thinking tokens, a small accuracy cost

Across twelve benchmarks, BottleCap reports an average of 37.2 per cent fewer thinking tokens. Accuracy comes out at 85.8 per cent against 86.6 per cent for the base model, so a little under one point is traded for roughly a third less output. These are the developer's own measurements. For anyone running an open reasoning model on their own hardware, shorter reasoning means faster answers and more requests handled per GPU.

Builds and licence terms

BottleCap publishes quantised builds in GGUF, FP8, NVFP4 and MLX formats alongside the full weights. The licence deserves a careful read: PolyForm Small Business 1.0.0 allows use by small businesses only, with a separate grant for personal use, and larger organisations must ask BottleCap for a commercial licence. The developer provides no hosted API.

TrustList summary

BottleCap AI has fine-tuned Alibaba's Qwen3.8-27B into ThinkingCap-Qwen3.8-27B, a version that reasons in fewer thinking tokens. Input can be text or images, output is text, and it serves a context of 262,144 tokens across 27.8 billion parameters.

What ThinkingCap-Qwen3.8-27B offers

  • Shorter reasoning: about 37.2 per cent fewer thinking tokens on average across twelve benchmarks, per the developer.
  • Accuracy: 85.8 per cent on those benchmarks, against 86.6 for the base model.
  • Quantised builds: GGUF, FP8, NVFP4 and MLX builds alongside the full weights.
  • Speed and throughput: fewer thinking tokens mean faster answers and more requests per GPU for teams running the model themselves.

Who it serves

Small businesses may use the model under the PolyForm Small Business licence, while larger organisations need a commercial licence from BottleCap.

Key facts

  • Licence: PolyForm Small Business 1.0.0, with a separate personal-use grant.
  • Context: 262,144 tokens served.
  • Parameters: 27.8 billion.
  • Hosted API: none from the developer.

Summary written by TrustList from the business's own published information, Nov 2026.

Benchmarks & AI stats

BenchmarkOfficialCommunity avg
No benchmark scores yet. Be the first to add one.

“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).

Sign in to add a benchmark score for this model.

ThinkingCap-Qwen3.8-27B launches

Request a demo or quote from ThinkingCap-Qwen3.8-27B

Protected by reCAPTCHA - Google Privacy Policy and Terms apply.

By sending, you agree we may share your request and contact details with the provider once you confirm your email.

Reviews

Write the first review of ThinkingCap-Qwen3.8-27B

Used it? Your experience helps other buyers decide.

Write a review

Questions & answers

No questions yet. Be the first to ask about ThinkingCap-Qwen3.8-27B.

ThinkingCap-Qwen3.8-27B alternatives

Similar listings buyers compare against this one.

Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems

Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks

Liquid AI's open-weight decision model, giving yes or no, choice or score answers from text and images

Open-weight translation model from Bilibili's Index team, covering 150 languages and following glossary and style instructions

Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0

Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure

Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms

Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio