Similarly, the 26B A4B Gemma 4 and the 35B A3B Qwen 3.6 identify it clearly, give me the title and trends analysis fairly accurately. While this 12B spits out gobbledygook about it having something to do with hard-drive capacity. It's like it can barely see, gets the very broad strokes (knows it's looking at some kind of chart), but can't identify any details clearly.

Even if it can't fully pass much, there are so many tests against most of the scenarios that you can get a fairly rich report beyond the pass@1 stat. See e.g. this DeepSWE report against the Minimax M3 model: https://entrpi.github.io/misc/deep-swe-minimax-m3/

  Qwen3.6-35B-A3B   vs   Claude Haiku 4.5
    reasoning mode · AA Intelligence Index v4.0
  
  46.0 ┤   ↖ better — cheaper · smarter · faster
       │
       │
  44.0 ┤     ╭─────╮
       │     │  ●  │ Qwen3.6-35B-A3B
       │     ╰─────╯
  42.0 ┤
       │
       │
  40.0 ┤
       │
       │
  38.0 ┤                                       ╭───╮
       │                      Claude Haiku 4.5 │ ○ │
       │                                       ╰───╯
  36.0 ┤
       └┬─────────┬─────────┬─────────┬─────────┬────────┬
        $200    $300      $400      $500      $600    $700
  
    x → cost to run the index (USD)        lower is better
    y → AA intelligence index              higher is better
  
    bubble area = output speed (tokens / sec)
          ╭─────╮                  ╭───╮
          │  ●  │ Qwen ~196 t/s    │ ○ │ Haiku ~93 t/s
          ╰─────╯                  ╰───╯
  
    ┌─────────────────────┬──────────┬──────────┬───────────┐
    │ model               │ AA index │ run cost │ out speed │
    ├─────────────────────┼──────────┼──────────┼───────────┤
    │ Qwen3.6-35B-A3B    ●│   43.5   │   $280   │  196 t/s  │
    │ Claude Haiku 4.5   ○│   37.1   │   $620   │   93 t/s  │
    └─────────────────────┴──────────┴──────────┴───────────┘


    COST PER TOKEN   ≠   COST PER TASK  
    output tokens per index run:
       Haiku 4.5    87.3M   (79.3M reasoning + 8.0M answer)
       Qwen3.6     143.2M   (131.7M reasoning + 11.5M answer)
       → Qwen emits 1.64× more output
  
    ── output speed (tokens / sec) ──────────  raw rate · higher = faster
       Qwen3.6     100%   ~196 t/s
       Haiku 4.5   ~47%   ~93 t/s
                                                  → Qwen ~2.1× faster per token
  
          ╎   1.64× more tokens  <  2.1× faster rate
          ▼
  
    ── solution speed (per finished answer) ──  higher = faster
       Qwen3.6     100%
       Haiku 4.5   ~78%
                                                  → Qwen ~1.3× FASTER to a solution
  
    SCORECARD
                            intelligence    cost / task     speed to solution
     Qwen3.6-35B-A3B        43.5            $280            ~1.3× faster 
     Claude Haiku 4.5       37.1            $620            (slower)
  
     → Qwen wins all three. The reasoning blow-up (1.64×) is smaller than
       the raw-speed edge (2.1×), so Qwen stays ahead per task.

We're just missing the establishment of a decorum of, "even if you do feel like you need to prompt the AI before responding, and even if you like the response, you still need to paraphrase and synthesize to avoid coming off rude and inhuman."

Mythos is an exception that's larger.

We know they serve the model on TPU 8i, which we have plenty of hard specs for (so we know the key constraints: total memory and bandwidth and compute flops). We can also set a ceiling on the compute complexity and memory demand of the model based on knowing they will be at least as efficient as what is disclosed in the Deepseek V4 Technical Report.

We can also assume that the model was explicitly built to run efficiently in a RadixAttention style batched serving scenario on a single TPU 8i (so no tensor parallelism, etc. to avoid unnecessary overheads... Google explicitly designed the 8th-generation inference architecture to eliminate the need for tensor sharding on mid-sized models).

We know Google intends to serve this model at a floor speed of around 280 tok/s too.

Putting all these pieces together, we can confidently say this model is ~250-300B total, and 10-16B active parameters. Likely mostly FP4 with FP8 where it matters most.

Visual:

  ┌────────────────────────────────────────────────────────┐
  │                   TPU 8i VRAM (288 GB)                 │
  ├───────────────────────────┬────────────────────────────┤
  │   Static Model Weights    │  Dynamic Allocations &     │
  │   (250B - 300B @ Mixed    │  Compressed KV Caches      │
  │   FP4/FP8)                │  (RadixAttention / SRAM)   │
  │   ~110 GB - 150 GB        │  ~138 GB - 178 GB          │
  └───────────────────────────┴────────────────────────────┘

Edit: There's one factor I under-rated in my initial estimate... TurboQuant. This is a compute to KV memory use tradeoff. It's plausible with TurboQuant at a quality-neutral setting they've gotten the model up to 400B with similar economics. This is a variable effecting concurrency and the the way they decided total model size was likely based on what they see for the average user's average KV cache depth in real-world usage.

I'm thinking something that could be a major upgrade in spirit to the long-in-the-tooth (released a decade ago) Casio FX-CG500.

Could use the soon to be released ARM C-1 Nano and Pro cores in an SoC with stacked 2GB LPDDR4, USB-C charging to a large battery, high-res transflective LCD...

Mockup "AxiomPad Pro X1": https://enia.cc/out/axiompad-cas-mock.png

Prompts like, "Please write a comprehensive report on _____ to work with _____. Include a holistic report on architecture and meaning and purpose of all involved systems. Describe the why and how of the changes in depth and include a full glossary of terms and systems. Write as a new .md in docs when you are sure there are no major gaps in your understanding. Include a report on the plan to _____."

Will be the rough shape you want to get it to dig through all the relevant code and make relevant architectural diagrams. Guide more or less towards specifics as appropriate. This has worked well since Opus 4.5.

The best tests are your own custom personal-task-relevant standardized tests (which the best models can't saturate, so aiming for less than 70% pass rate in the best case).

All this is to say that most people are not doing the latter and their vibes are heavily confounded to the point of being mostly meaningless.