Microgpt in pure C hits 10M tps on Apple m5

A new open-source project, microgpt, implements a character-level GPT in a single dependency-free C file, achieving 10,168,430 tokens per second on an Apple M5 Pro with NEON and 6,927,775 tokens per second on an AMD Ryzen 5 5600H with AVX2. The model, with 4,192 parameters, trains on 20,000 of 32,033 names in seconds and generalizes to unseen data, scoring 2.2039 nats per character on held-out names.

The most atomic way to train and inference a GPT in pure, dependency-free C. A character-level transformer with forward pass, backprop, Adam and sampling, in one C file with nothing beyond libc. It trains on ~32k names in a couple of seconds and generates new ones. make run Or run it directly, on any corpus with one item per line: ./microgpt data/names.txt Builds on macOS, Linux and Windows MSYS2 , on ARM64 with NEON and x86-64 with AVX2. The Makefile picks the flags for the host. step 5000 / 20000 | loss 2.6036 avg 2.2940 step 10000 / 20000 | loss 1.9639 avg 2.2564 step 15000 / 20000 | loss 2.7007 avg 2.2151 step 20000 / 20000 | loss 2.3463 avg 2.2201 inference sample 1: kayley sample 2: maria sample 3: arana sample 4: shayan sample 5: jayden sample 6: saria sample 7: kaylen sample 8: amari sample 9: alina sample 10: mailyn c fp32+NEON 10168430 tok/sec The model has 4192 parameters and generalises rather than memorises. Trained on 20000 of the 32033 names, it scores 2.2054 nats per character on those and 2.2039 on the 12033 it never saw, beating an interpolated trigram that has nearly five times as many parameters. Training and inference use separate forward passes. gpt forward stores activations for backprop; gpt forward infer is a specialised single-token path whose logits match it to within fp32 rounding. docs/PERFORMANCE.md /vixhal-baraiya/microgpt-c/blob/main/docs/PERFORMANCE.md covers how that path works and what limits it. | machine | backend | tok/sec | |---|---|---| | Apple M5 Pro | NEON | 10,168,430 | | AMD Ryzen 5 5600H | AVX2 | 6,927,775 |