Core AI model zoo

Agents-A1-4B — Apple Core AI (.aimodel)

The source card describes the model as “Agents‑A1 is a long-horizon agentic model from InternScience”. Its declared license is apache-2.0. This bundle converts the TEXT tower of the Qwen3_5ForConditionalGeneration checkpoint. Its text_config equals Qwen/Qwen3.5-4B, so the zoo’s 4B recipe applies unchanged with an HF-id swap.

[!NOTE] Exported with coreai-core 1.0.0b2 for the Mac-class coreai-pipelined GPU engine on macOS 27. The canonical bundle path is gpu-pipelined-b2/agents_a1_4b_decode_int8hu_block32_sym/.

Bundles

Measured

Apple M4 Max GPU, macOS 27.0, Release tools. Correctness uses greedy coreai-pipelined, COREAI_CHUNK_THRESHOLD=1, and warmup off; the reference oracle runs fp32 on CPU.

Check Result
France raw prompt: The capital of France is 16/16 tokens exact against fp32 CPU (transcript)
Alphabet raw prompt: The alphabet begins A, B, C, D, E, F, 16/16 tokens exact against fp32 CPU (transcript)
Chat no-think, 1+1=? 1 + 1 = 2; 8/8 IDs exact including EOS 248046 (the transcript records the 7 tokens before EOS)
Tool calls, no-think 3/3 in the template’s XML format: get_weather(Tokyo, celsius), convert_currency(250, USD, EUR), search_web("Swift 6.3 release notes"); stopped at 40 / 55 / 31 tokens, with nothing after each closing tag
Swift-side template Template applied without error; </think> followed by Canberra; stopped at 201 tokens
Eight fixed sanity questions, no-think 8/8 expected-substring hits and 8/8 EOS stops
First / cached load (llm-runner) 3.377 s / 0.807 s

Decode speed, llm-benchmark Release, -p 128 -g 256 -n 3, no other GPU job on the machine (M4 Max, macOS 27.0 26A428):

Trial Prompt tok/s Generation tok/s
1 82.1 72.3
2 82.9 72.3
3 82.9 72.4
Mean 82.6 72.3

Prompt tok/s is the pipelined S=1 prefill (COREAI_CHUNK_THRESHOLD=1), which is why it sits near the decode rate.

Run (macOS)

Use the zoo’s engine patch stack: apps/coreai-shared-product.patch followed by apps/coreai-pipelined-extra-states.patch. From the downloaded gpu-pipelined-b2/ directory, run a Release build:

COREAI_CHUNK_THRESHOLD=1 llm-benchmark --model agents_a1_4b_decode_int8hu_block32_sym -p 128 -g 256 -n 3

The original chat template is preserved:

<tool_call>
<function=…>
<parameter=…>…</parameter>
</function>
</tool_call>

iPhone

No iPhone bundle is provided. The 4B-class graph exceeds on-device GPU specialization and needs ahead-of-time (h18p) compilation.

Reproduce

The unchanged exporter and recipe specify the conversion:

python3 conversion/zoo_convert.py show agents-a1-4b
python3 conversion/export_qwen3_5_decode_pipelined.py int8hu --head-sym --hf-id InternScience/Agents-A1-4B

Gate transcripts: gate-agents-a1-4b.json, gate-agents-a1-4b-alphabet.json, and gate-agents-a1-4b-chat-nothink.json. Runtime and quantization details are in the zoo’s pipelined-engine notes.


⬇️ Download: 🤗 mlboydaisuke/Agents-A1-4B-CoreAI — this card and the model page are the same document. Reproduce it: python3 conversion/zoo_convert.py show agents-a1-4b prints the command and its prerequisites; recipe.toml is the record. Conversion and gates were run by Codex (gpt-6-astra) under supervision; the transcripts above are its evidence.