ZMIME
Comparison · 3 models · Updated Nov 18, 2026

GPT-5.6 Luna vs Inkling Small vs Qwen3.8 27B

GPT-5.6 Luna comes out ahead, 80 to 75 and 73 on our weighted score, and it is the cheaper option too.

Price in rating: Excluded · Include price

  1. Our pick

    OpenAI

    GPT-5.6 Luna

    Released Jul 9, 2026

    80/100
    • ECI156.7
    • Price$0.20 / $1.20
    • Context1.05M
  2. Thinking Machines

    Inkling Small

    Released Jul 30, 2026

    75/100
    • ECI150.8
    • Price$0.58 / $1.44
    • Context1.05M
  3. Alibaba (Qwen)

    Qwen3.8 27B

    Released Aug 14, 2026

    73/100
    • ECI152.7
    • Price$0.40 / $2.42
    • Context262K
01 — Verdict

GPT-5.6 Luna is our pick

GPT-5.6 Luna is the better all-round choice, scoring 80/100 against Inkling Small (75) and Qwen3.8 27B (73). It leads on capability and inputs & features. The score weighs capability 67%, inputs & features 20%, context window 13%.

  • CapabilityGPT-5.6 LunaCapabilities Index (ECI): GPT-5.6 Luna 156.7 · Qwen3.8 27B 152.7 · Inkling Small 150.8
  • Lowest priceGPT-5.6 LunaGPT-5.6 Luna $0.45 · Inkling Small $0.795 · Qwen3.8 27B $0.905 per 1M tokens (3:1 blend)
  • Longest contextGPT-5.6 Luna and Inkling SmallGPT-5.6 Luna 1,050,000 · Inkling Small 1,048,576 · Qwen3.8 27B 262,144 tokens
  • Widest inputsSame inputsGPT-5.6 Luna: Text, Images, PDFs · Inkling Small: Text, Images, Audio · Qwen3.8 27B: Text, Images, Video
  • Self-hostingInkling Small and Qwen3.8 27BPublishes downloadable weights (Apache-2.0)
How the score is built
MeasureWeightGPT-5.6 LunaInkling SmallQwen3.8 27B
CapabilityCapabilities Index (ECI)67%847880
Inputs & features20%807070
Context window13%656543
Overall100%80/10075/10073/100
02 — Side by side

Every spec in one table

Highlighted cells lead their row. Dashes mean the data is not published.

GPT-5.6 Luna vs Inkling Small vs Qwen3.8 27B specifications side by side
SpecificationGPT-5.6 LunaOpenAIInkling SmallThinking MachinesQwen3.8 27BAlibaba (Qwen)
Capability
Capabilities Index (ECI)156.7 (best)150.8152.7
ECI rank#36 of 158 (best)#65 of 158#57 of 158
GPQA DiamondGraduate-level science questions91.6% (best)88.5%—
FrontierMath Tiers 1–3Research-level mathematics82.1% (best)46.3%—
OTIS Mock AIME 2024–2025Competition mathematics98.3% (best)90.0%—
SimpleQA VerifiedShort factual questions41.0% (best)19.1%—
Price per million tokens
Input$0.20 (best)$0.58$0.40
Output$1.20 (best)$1.44$2.42
Cached input$0.02 (best)$0.116—
Blended (3:1)$0.45 (best)$0.795$0.905
Long-context rateOver 272K: $0.40 / $1.80Same rateSame rate
Price sourceOfficial OpenAI APIOfficial Thinking Machines APIMedian of 67 providers
Limits
Context window1,050,000 tokens (best)1,048,576 tokens262,144 tokens
Max output128,000 tokens1,048,576 tokens (best)32,768 tokens
Inputs and features
TextYesYesYes
ImagesYesYesYes
PDFsYesNoNo
AudioNoYesNo
VideoNoNoYes
ReasoningYeslow · medium · high · xhigh · maxYeslow · medium · high · xhigh · maxYes
Tool callingYesYesYes
Structured outputYesNoNo
Availability
WeightsProprietaryOpenApache-2.0Open
API model IDgpt-5.6-lunathinkingmachines/Inkling-Small:peft:262144—
API providers571272 (best)
ReleasedJul 9, 2026Jul 30, 2026Aug 14, 2026
Knowledge cutoffFeb 16, 2026——
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • GPT-5.6 Luna$4.40
  • Inkling Small$8.68
  • Qwen3.8 27B$8.84
04 — Questions

Which should you choose?

Which is better: GPT-5.6 Luna, Inkling Small or Qwen3.8 27B?

GPT-5.6 Luna is the better all-round choice, scoring 80/100 against Inkling Small (75) and Qwen3.8 27B (73). It leads on capability and inputs & features. The score weighs capability 67%, inputs & features 20%, context window 13%.

Which is cheaper, GPT-5.6 Luna, Inkling Small or Qwen3.8 27B?

GPT-5.6 Luna is cheaper at $0.20 input / $1.20 output per million tokens (official OpenAI API price). Inkling Small costs $0.58 input / $1.44 output per million tokens (official Thinking Machines API price); Qwen3.8 27B costs $0.40 input / $2.42 output per million tokens (median across 67 API providers). At a typical mix of three input tokens to one output token, that is $0.45 per million tokens for GPT-5.6 Luna versus $0.795 for Inkling Small (1.8× as much) and $0.905 for Qwen3.8 27B (2× as much).

Which scores higher on benchmarks?

GPT-5.6 Luna scores higher on the Capabilities Index (ECI): GPT-5.6 Luna 156.7 (#36 of 158), Qwen3.8 27B 152.7 (#57 of 158) and Inkling Small 150.8 (#65 of 158). Their confidence ranges do not overlap (154.9–158.8 vs 150.7–154.8), so the gap is a real one.

Which is better for coding?

There are no published SWE-bench Verified results for GPT-5.6 Luna, Inkling Small and Qwen3.8 27B yet, so there is no like-for-like coding score. On overall capability, GPT-5.6 Luna leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.

Which has the bigger context window?

GPT-5.6 Luna and Inkling Small have the largest context windows (1,050,000 and 1,048,576 tokens), against 262,144 for Qwen3.8 27B. Maximum output per response: GPT-5.6 Luna up to 128,000, Inkling Small up to 1,048,576, Qwen3.8 27B up to 32,768 tokens.

Which can read images, PDFs, audio or video?

GPT-5.6 Luna accepts text, images and PDFs; Inkling Small accepts text, images and audio; Qwen3.8 27B accepts text, images and video. They handle the same number of input types.

Are any of these open source?

Inkling Small and Qwen3.8 27B publishes its weights (Apache-2.0) and can be self-hosted; GPT-5.6 Luna is proprietary.

Which is newer?

Qwen3.8 27B is the newest, released Aug 14, 2026. Inkling Small came out Jul 30, 2026; GPT-5.6 Luna came out Jul 9, 2026. Knowledge cutoff: GPT-5.6 Luna Feb 16, 2026.

How do you decide the winner?

This rating uses capability (67%, independent benchmark results); inputs & features (20%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (13%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 18, 2026.