Kalkulator
Kalkulator Biaya Fine-tuning LLM
Hitung biaya fine-tuning — token training × tarif per juta, plus uplift per token untuk inference model custom.
Harga diperbarui:
AITOT LLM Fine-tuning Cost calculator memperkirakan biaya training + inference uplift untuk model fine-tuned di OpenAI (GPT-4o, GPT-4o-mini, o3), Anthropic Claude (invite-only), Google Vertex (Gemini), Together AI (LoRA untuk Llama 4, Qwen, Mistral).
Training cost = training tokens × epochs × per-million rate. OpenAI GPT-4o-mini: $3/M training tokens. Together Llama 4 70B LoRA: $1.20/M. Mayoritas fine-tunes produksi $50-$500 one-time.
Toggle epochs (default 3) dan volume inference. Di bawah 10M token/bulan, fine-tuning jarang kalahkan prompts. Di atas 100M, fine-tuned smaller model kalahkan larger model dengan prompts 3-10×.
Total tahun 1 · termurah
Fireworks · Llama 4 8B
$248
| Provider | Model dasar | Biaya training | Inference bulanan | Total tahun 1 |
|---|---|---|---|---|
| Fireworks | Llama 4 8B ≤16B LoRA SFT tier | $8 | $20 | $248 |
| Cohere | Command R | $30 | $48 | $606 |
| OpenAI | GPT-4o mini Stale — OpenAI moved to per-hour training 2026-05; verify pending | $45 | $48 | $621 |
| Mistral | Mistral Small 3 $2/mo hosting per deployed adapter | $45 | $58 | $741 |
| Fireworks | Llama 4 70B 16-80B LoRA SFT tier | $45 | $90 | $1,125 |
| Together | Llama 3.3 70B Legacy v3 line; verify pending 2026-05-18 — no longer top-listed on Together pricing | $75 | $88 | $1,131 |
| OpenAI | GPT-5 mini Stale — OpenAI moved to per-hour training 2026-05; verify pending | $60 | $96 | $1,212 |
| Together | Llama 4 Maverick (LoRA SFT) $16 minimum charge; Maverick = ~70B-class | $120 | $120 | $1,560 |
| OpenAI | o3-mini Stale — OpenAI moved to per-hour training 2026-05; verify pending | $75 | $136 | $1,707 |
| Together | Llama 4 Maverick (LoRA DPO) | $300 | $120 | $1,740 |
| AWS Bedrock | Claude Haiku 4.5 (custom) Provisioned throughput required | $120 | $303 | $3,756 |
| Mistral | Mistral Large 2 | $135 | $564 | $6,903 |
| OpenAI | GPT-4o Stale — OpenAI moved to per-hour training 2026-05; verify pending | $375 | $600 | $7,575 |
Biaya training = token × epoch × tarif per juta. Inference memakai tarif fine-tuned yang selalu lebih tinggi dari model dasar. Total tahun 1 = training sekali + 12 bulan inference.
Yang dilakukan kalkulator ini
Multi-provider
OpenAI fine-tuning, Together LoRA, Vertex tuning, self-host estimates.
Training + inference split
Training cost one-time dipisah dari uplift bulanan.
Slider epochs
Default 3 epochs.
Modeling inference uplift
Fine-tuned biaya 1.5-3× base.
Total tahun-1
Training one-time + 12 bulan inference = satu angka budget.
LoRA vs full fine-tuning
LoRA Together 10× lebih murah dari full FT OpenAI.
Perbandingan cepat
Fine-tuning 5M training tokens, 50M inference/bulan, 3 epochs
| Provider | Training Cost | Inference Uplift | Total Tahun-1 |
|---|---|---|---|
| Together Llama 4 70B (LoRA) | $18 | +$50/bulan | $618 |
| OpenAI GPT-4o-mini | $45 | +$120/bulan | $1,485 |
| Google Gemini 2.5 Flash tune | $75 | +$150/bulan | $1,875 |
| OpenAI GPT-4o | $375 | +$1,200/bulan | $14,775 |
| OpenAI o3 | $2,250 | +$3,500/bulan | $44,250 |
Tahun-1 = training + 12 × uplift bulanan.
Cara menggunakan kalkulator
Hitung training + inference uplift cost untuk LLMs fine-tuned.
- 1
Masukkan training tokens
Total token di dataset training. 100 examples × 500 tokens = 50k.
- 2
Set epochs
Default 3. Lebih dari 4 biasanya overfit.
- 3
Perkirakan inference bulanan
Berapa token fine-tuned model akan serve/bulan.
- 4
Bandingkan provider
LoRA Together termurah; OpenAI full FT termahal.
Kenapa pakai kalkulator ini
- ✓5 provider diperbarui bulanan
- ✓Training + inference split
- ✓Perbandingan LoRA vs full FT
- ✓Angka budget tahun-1
- ✓Modeling epoch + token
- ✓Tanpa login