ZMIME
Comparison · 3 models · Updated Nov 18, 2026

Qwen3.8 2.4T A95B vs GPT-6.2 Astra vs Fugu Max

No independent capability benchmark covers every model here yet. The score uses price, inputs & features and context window.

Price in rating: Included · Exclude price

  1. Alibaba (Qwen)

    Qwen3.8 2.4T A95B

    Released Aug 12, 2026

    32/100
    • ECI—
    • Price$2.00 / $6.00
    • Context262K
  2. OpenAI

    GPT-6.2 Astra

    Released Nov 17, 2026New

    37/100
    • ECI174.2
    • Price$8.00 / $40.00
    • Context1.05M
  3. Sakana AI

    Fugu Max

    Released Sep 11, 2026

    47/100
    • ECI—
    • Price$2.00 / $6.00
    • Context1M
01 — Verdict

Capability evidence is incomplete

There is not enough shared capability evidence to name an overall winner. The scores (Fugu Max 47/100, GPT-6.2 Astra 37/100, Qwen3.8 2.4T A95B 32/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

  • CapabilityNot enough dataNo independent benchmark covers every model here yet
  • Lowest priceQwen3.8 2.4T A95B and Fugu MaxQwen3.8 2.4T A95B $3.00 · Fugu Max $3.00 · GPT-6.2 Astra $16.00 per 1M tokens (3:1 blend)
  • Longest contextGPT-6.2 AstraGPT-6.2 Astra 1,050,000 · Fugu Max 1,000,000 · Qwen3.8 2.4T A95B 262,144 tokens
  • Widest inputsGPT-6.2 Astra and Fugu MaxQwen3.8 2.4T A95B: Text · GPT-6.2 Astra: Text, Images, PDFs · Fugu Max: Text, Images, PDFs
  • Self-hostingQwen3.8 2.4T A95BPublishes downloadable weights (qwen3.8-max)
How the score is built
MeasureWeightQwen3.8 2.4T A95BGPT-6.2 AstraFugu Max
Price50%26026
Inputs & features30%358070
Context window20%436564
Specs & price only100%32/10037/10047/100

Left out because at least one model lacks the data: capability. The remaining weights were rescaled.

02 — Side by side

Every spec in one table

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

Qwen3.8 2.4T A95B vs GPT-6.2 Astra vs Fugu Max specifications side by side
SpecificationQwen3.8 2.4T A95BAlibaba (Qwen)GPT-6.2 AstraOpenAIFugu MaxSakana AI
Capability
Capabilities Index (ECI)—174.2—
ECI rank—#1 of 158—
GPQA DiamondGraduate-level science questions—96.8%—
FrontierMath Tiers 1–3Research-level mathematics—92.6%—
OTIS Mock AIME 2024–2025Competition mathematics—100%—
SimpleQA VerifiedShort factual questions—78.5%—
Price per million tokens
Input$2.00 (best)$8.00$2.00 (best)
Output$6.00 (best)$40.00$6.00 (best)
Cached input—$0.40—
Blended (3:1)$3.00 (best)$16.00$3.00 (best)
Long-context rateSame rateOver 272K: $16.00 / $60.00Same rate
Price sourceMedian of 26 providersOfficial OpenAI APIMedian of 9 providers
Limits
Context window262,144 tokens1,050,000 tokens (best)1,000,000 tokens
Max output131,072 tokens (best)128,000 tokens128,000 tokens
Inputs and features
TextYesYesYes
ImagesNoYesYes
PDFsNoYesYes
AudioNoNoNo
VideoNoNoNo
ReasoningYesYeslow · medium · high · xhigh · maxYes
Tool callingYesYesYes
Structured outputNoYesNo
Availability
WeightsOpenqwen3.8-maxProprietaryProprietary
API model ID—gpt-6.2-astra—
API providers26 (best)129
ReleasedAug 12, 2026Nov 17, 2026Sep 11, 2026
Knowledge cutoff—May 18, 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.

  • Qwen3.8 2.4T A95B$32.00
  • GPT-6.2 Astra$160.00
  • Fugu Max$32.00
04 — Questions

Which should you choose?

Which is better: Qwen3.8 2.4T A95B, GPT-6.2 Astra or Fugu Max?

There is not enough shared capability evidence to name an overall winner. The scores (Fugu Max 47/100, GPT-6.2 Astra 37/100, Qwen3.8 2.4T A95B 32/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

Which is cheaper, Qwen3.8 2.4T A95B, GPT-6.2 Astra or Fugu Max?

Qwen3.8 2.4T A95B is cheaper at $2.00 input / $6.00 output per million tokens (median across 26 API providers). Fugu Max costs $2.00 input / $6.00 output per million tokens (median across 9 API providers); GPT-6.2 Astra costs $8.00 input / $40.00 output per million tokens (official OpenAI API price). At a typical mix of three input tokens to one output token, that is $3.00 per million tokens for Qwen3.8 2.4T A95B versus $3.00 for Fugu Max (1× as much) and $16.00 for GPT-6.2 Astra (5.3× as much).

Which scores higher on benchmarks?

There is no independent benchmark that covers all three models yet. Qwen3.8 2.4T A95B has not been scored yet, GPT-6.2 Astra has an ECI of 174.2 and Fugu Max has not been scored yet.

Which is better for coding?

There are no published SWE-bench Verified results for Qwen3.8 2.4T A95B, GPT-6.2 Astra and Fugu Max yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.

Which has the bigger context window?

GPT-6.2 Astra has the largest context window at 1,050,000 tokens, against 1,000,000 for Fugu Max and 262,144 for Qwen3.8 2.4T A95B. Maximum output per response: Qwen3.8 2.4T A95B up to 131,072, GPT-6.2 Astra up to 128,000, Fugu Max up to 128,000 tokens.

Which can read images, PDFs, audio or video?

Qwen3.8 2.4T A95B accepts text; GPT-6.2 Astra accepts text, images and PDFs; Fugu Max accepts text, images and PDFs. GPT-6.2 Astra handles the widest range of inputs.

Are any of these open source?

Qwen3.8 2.4T A95B publishes its weights (qwen3.8-max) and can be self-hosted; GPT-6.2 Astra and Fugu Max is proprietary.

Which is newer?

GPT-6.2 Astra is the newest, released Nov 17, 2026. Fugu Max came out Sep 11, 2026; Qwen3.8 2.4T A95B came out Aug 12, 2026. Knowledge cutoff: GPT-6.2 Astra May 18, 2026.

How do you decide the winner?

This rating uses price (50%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (30%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (20%, 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.