GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Ajay0007
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Ajay0007
Model Tree
Qwen/Qwen3-0.6B
Input Modalities
Output Modalities
Supported Functionality
text
photo ──► on-device OCR (ML Kit / Apple Vision, ~0.3 s) ──► this model (~1 s) ──► JSON ──► sum check ──► split UI └─ fails → ask the user to reviewThe sum check (do item prices + tax + service − discount equal the total?) is a strong confidence signal: in our
tests, receipts that pass it were fully correct ~89% of the time, versus ~25% for receipts that fail it.
receipt_utils.sum_check() in this repo implements it.
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom huggingface_hub import hf_hub_downloadimport importlib.util, sys repo = "Ajay0007/receipt-extractor-qwen3-0.6b"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo) # helper functions shipped with the model (prompt, OCR line grouping, JSON parsing, sum check)spec = importlib.util.spec_from_file_location("receipt_utils", hf_hub_download(repo, "receipt_utils.py"))ru = importlib.util.module_from_spec(spec); spec.loader.exec_module(ru) ocr_text = """1 EGG TART 13,0002 CHOCO CUSTARD PASTRY 24,000REDBEAN BREAD 9,000TOTAL 46,000CASH 50,000CHANGE 4,000""" messages = [{"role": "user", "content": ru.build_prompt(ocr_text)}]inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_dict=True, return_tensors="pt", enable_thinking=False)out = model.generate(**inputs, max_new_tokens=768, do_sample=False)receipt = ru.parse_output(tok.decode(out[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))print(receipt)print("sum check passed:", ru.sum_check(receipt))Output for the example above:
json
{"items": [{"name": "EGG TART", "qty": "1", "price": "13,000"}, {"name": "CHOCO CUSTARD PASTRY", "qty": "2", "price": "24,000"}, {"name": "REDBEAN BREAD", "qty": "1", "price": "9,000"}], "total": "46,000"}Use the prompt exactly as build_prompt writes it. The model was trained on that wording. Use greedy decoding
(do_sample=False). For Qwen3, keep enable_thinking=False (the model was trained without a thinking block).
The model expects one visual receipt line per row, with the tokens on a line ordered left to right (so an item
name and its price sit on the same row). Raw OCR engines return boxes, not lines. Use
receipt_utils.ocr_boxes_to_text([(x_min, y_min, y_max, text), ...]), which is the same grouping used to build the
training data. Port this ~20-line function to Kotlin/Swift for your app.
model.litertlm is an int8 build (weights int8, activations fp32) made with
litert-torch (dynamic_wi8_afp32, prefill 256, KV cache 1024 tokens).
This build was verified to produce correct JSON in the LiteRT-LM runtime (4/5 smoke-test receipts fully correct); a full accuracy/latency benchmark of the int8 build was not completed.
python
import litert_lm # pip install litert-lm-apifrom huggingface_hub import hf_hub_downloadpath = hf_hub_download("Ajay0007/receipt-extractor-qwen3-0.6b", "model.litertlm")engine = litert_lm.Engine(path, backend=litert_lm.Backend.CPU(thread_count=4), max_num_tokens=2048)conv = engine.create_conversation(sampler_config=litert_lm.SamplerConfig(top_k=1, temperature=0.0), max_output_tokens=768, chat_template=open(hf_hub_download("Ajay0007/receipt-extractor-qwen3-0.6b", "litert_chat_template.jinja")).read())reply = conv.send_message(ru.build_prompt(ocr_text))print(reply["content"] if isinstance(reply, dict) else reply)For apps, load the same file with the LiteRT-LM Kotlin API (Android, stable) or Swift API (iOS), or use the Google AI Edge Gallery app to try it on a phone and see its speed.
Important for Qwen3: pass the included litert_chat_template.jinja as the conversation's chat template (as above). The template bundled by the converter adds a system prompt that makes the model ignore the task.
| Dataset | Receipts used | Region / language | Labels used |
|---|---|---|---|
| CORD-v2 (CC-BY-4.0), train split | 800 | Indonesia (Indonesian/English, IDR) | items, priced add-ons, subtotal, tax, service, discount, total |
| WildReceipt, train split | 1,013 | mostly US/EU photos (English, some German/French) | items, subtotal, tax, tip→service, total |
| Total | 1,813 |
Inputs were produced by running RapidOCR (PP-OCR) on the images, so the model learned from realistic, noisy OCR, not clean ground-truth text. CORD labels were corrected: priced add-ons stored as sub-items in CORD (e.g. "Extra Jelly 5,000") were restored, and a few annotation errors fixed. Store names, addresses, dates and payment lines (cash, change, card) are deliberately not extracted.
Full fine-tune (all weights) with plain supervised next-token loss on the assistant JSON only (prompt tokens masked).
| Setting | Value |
|---|---|
| Epochs | 5 |
| Learning rate / schedule | 5e-5, cosine, 5% warmup |
| Batch size | 8 |
| Max sequence length | 2,048 tokens |
| Precision | bf16, gradient checkpointing |
| Hardware / time | 1× NVIDIA A40, ~17 min |
| Target format | compact JSON, null fields dropped (shorter output = faster on device) |
We also tested LoRA (r=16, all linear layers, lr 2e-4): it matched the full fine-tune on CORD (70% vs 72% split-ready) and is a good option if you want small swappable adapters.
All numbers use OCR from RapidOCR, greedy decoding and the bf16 weights unless noted. Split-ready = every line
item (name + price) and the total are exactly right, so the bill can be split with no edits. Item F1 = item
matched on name (fuzzy ≥ 80) and exact price. Amounts are compared by value (60.000 = 60,000, 525 = 525.00).
| Test set | Item F1 | Total correct | Split-ready |
|---|---|---|---|
| CORD-v2 test+val, corrected labels (200, Indonesia) | 85.9% | 94.3% | 75.5% |
| WildReceipt test (472, US/EU photos) | 59.9% | 91.2% | 37.3% |
| SROIE test (347, Malaysia, total only) — unseen dataset | – | 86.2% | – |
| Real-world photos, Latin script (30, Wikimedia Commons, hand-verified) — unseen | 47.9% | 56.7% | 36.7% |
| Real-world photos, CJK/Thai/Cyrillic (33, VLM-labeled) — unseen | 35.7% | 75.8% | 21.2% |
| Model | Trained on | CORD split / total | WildReceipt split / total | SROIE total | Photos split / total | Size |
|---|---|---|---|---|---|---|
| Donut-base CORD-v2 (public, image→JSON) | CORD | 82 / 98 | 1 / 54 | 31 | 0 / 13 | ~200M params |
| Gemma 3 270M | CORD | 64 / 93 | 4 / 48 | 33 | 0 / 20 | ~290 MB |
| LFM2.5-350M | CORD | 73 / 94 | 2 / 40 | 19 | 0 / 13 | 364 MB |
| Qwen3-0.6B | CORD | 76 / 93 | 10 / 64 | 49 | 3 / 47 | 614 MB |
| LFM2-VL-450M (image→JSON) | CORD | 74 / 92 | 3 / 52 | 43 | 7 / 47 | ~450M params |
| LFM2.5-350M — ours | CORD-fixed + WildReceipt | 78 / 96 | 31 / 88 | 85 | 13 / 50 | 364 MB |
| Qwen3-0.6B — ours | CORD-fixed + WildReceipt | 76 / 94 | 37 / 91 | 86 | 37 / 57 | 614 MB |
| LFM2-350M-Extract, zero-shot | — | 2 / 34 | 2 / 16 | 26 | – / – | 364 MB |
Key takeaways:
receipt_utils.to_number).Improve it for your users: fine-tune further on a few hundred labeled receipts from your target countries. In our experiments, adding diverse data mattered far more than model size.
Apache-2.0 (same as the base model). Fine-tuning data: CORD-v2 (CC-BY-4.0) and WildReceipt. This repository contains modified weights
(fine-tuned) of Qwen/Qwen3-0.6B.
photo ──► on-device OCR (ML Kit / Apple Vision, ~0.3 s) ──► this model (~1 s) ──► JSON ──► sum check ──► split UI └─ fails → ask the user to reviewfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom huggingface_hub import hf_hub_downloadimport importlib.util, sys repo = "Ajay0007/receipt-extractor-qwen3-0.6b"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo) # helper functions shipped with the model (prompt, OCR line grouping, JSON parsing, sum check)spec = importlib.util.spec_from_file_location("receipt_utils", hf_hub_download(repo, "receipt_utils.py"))ru = importlib.util.module_from_spec(spec); spec.loader.exec_module(ru) ocr_text = """1 EGG TART 13,0002 CHOCO CUSTARD PASTRY 24,000REDBEAN BREAD 9,000TOTAL 46,000CASH 50,000CHANGE 4,000""" messages = [{"role": "user", "content": ru.build_prompt(ocr_text)}]inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_dict=True, return_tensors="pt", enable_thinking=False)out = model.generate(**inputs, max_new_tokens=768, do_sample=False)receipt = ru.parse_output(tok.decode(out[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))print(receipt)print("sum check passed:", ru.sum_check(receipt)){"items": [{"name": "EGG TART", "qty": "1", "price": "13,000"}, {"name": "CHOCO CUSTARD PASTRY", "qty": "2", "price": "24,000"}, {"name": "REDBEAN BREAD", "qty": "1", "price": "9,000"}], "total": "46,000"}import litert_lm # pip install litert-lm-apifrom huggingface_hub import hf_hub_downloadpath = hf_hub_download("Ajay0007/receipt-extractor-qwen3-0.6b", "model.litertlm")engine = litert_lm.Engine(path, backend=litert_lm.Backend.CPU(thread_count=4), max_num_tokens=2048)conv = engine.create_conversation(sampler_config=litert_lm.SamplerConfig(top_k=1, temperature=0.0), max_output_tokens=768, chat_template=open(hf_hub_download("Ajay0007/receipt-extractor-qwen3-0.6b", "litert_chat_template.jinja")).read())reply = conv.send_message(ru.build_prompt(ocr_text))print(reply["content"] if isinstance(reply, dict) else reply)