gpt-oss-120b on a B200
Yes, you can fine-tune gpt-oss-120b on one B200 (180 GB) with QLoRA. 4-bit inference: yes, 65 GB of weights, room for 128K tokens of context or 621 concurrent 4K-token sequences.
Verdict by method
| Fine-tuning | Memory at 2K tokens | One B200? | Longest sequence (batch 1) | Largest batch at 2K | Largest LoRA rank |
|---|---|---|---|---|---|
| QLoRA (4-bit base) | 66 GB | Yes | 32K tokens | 24 | 1024 |
| LoRA (16-bit base) | 238 GB | No | — | — | — |
| Full fine-tuning | 1,873 GB | No | — | — | n/a |
| Inference | Memory at 8K | One B200? | Longest context | Concurrent 4K sequences |
|---|---|---|---|---|
| 16-bit (BF16/FP16) | 234 GB | No | — | — |
| 8-bit | 117 GB | Yes | 128K tokens | 290 |
| 4-bit | 66 GB | Yes | 128K tokens | 621 |
- QLoRA: yes, about 66 GB at 2K tokens, batch 1, rank 16; up to 32K-token sequences at batch 1 or batch 24 at 2K, LoRA rank up to 1024.
- LoRA: no, it needs about 238 GB at 2K tokens, more than the 180 GB available.
- Full fine-tuning: no, it needs about 1,873 GB at 2K tokens, more than the 180 GB available.
Where the memory goes
| Method | Weights | Training state | Activations | Total | On one B200 | 80 GB GPUs needed |
|---|---|---|---|---|---|---|
| QLoRA (4-bit base) | 62 GB | 0.2 GB | 4.0 GB | 66 GB | Fits | 1 |
| LoRA (16-bit base) | 234 GB | 0.2 GB | 4.0 GB | 238 GB | Doesn't fit | 4 |
| Full fine-tuning | 234 GB | 1,636 GB | 4.0 GB | 1,873 GB | Doesn't fit | 27 |
| Precision | Weights | KV cache at 8K | Total | At 128K context | On one B200 |
|---|---|---|---|---|---|
| 16-bit (BF16/FP16) | 234 GB | 0.3 GB | 234 GB | 238 GB | Doesn't fit |
| 8-bit | 117 GB | 0.3 GB | 117 GB | 122 GB | Fits |
| 4-bit | 65 GB | 0.3 GB | 66 GB | 70 GB | Fits |
The cheapest way to run it
Renting a B200 costs $7.99 an hour on Runpod Secure Cloud ($5.98 on Runpod Community Cloud). QLoRA on it processes about 34,412 tokens a second by our estimate, so 10 million training tokens take 5 minutes and cost about $0.64.
| Job (default settings) | Memory | Cheapest rental that fits | Per hour |
|---|---|---|---|
| QLoRA fine-tuning | 66 GB | A100 80GB | $2.19 |
| LoRA fine-tuning | 238 GB | 6 × L40S | $6.54 |
| 4-bit inference | 66 GB | A100 80GB | $2.19 |
| 16-bit inference | 234 GB | 6 × L40S | $6.54 |
On-demand rates seen on 15 November 2026: Runpod Secure Cloud, Lambda for the A100 40GB. Consumer cards count as single GPUs only; data-centre GPUs up to eight per node.
Try other settings
| Job | Memory | B200 (180 GB) | Smallest single GPU |
|---|---|---|---|
| QLoRA (4-bit base)2K tokens × batch 1, rank 16 | 66 GB | Fits | A100 80GB (80 GB) |
| LoRA (16-bit base)2K tokens × batch 1, rank 16 | 238 GB | 2 GPUs | None: needs several |
| Full fine-tuning2K tokens × batch 1 | 1,873 GB | 12 GPUs | None: needs several |
| Inference, 16-bit (BF16/FP16)8K-token context × 1 sequence | 234 GB | 2 GPUs | None: needs several |
| Inference, 8-bit8K-token context × 1 sequence | 117 GB | Fits | H200 (141 GB) |
| Inference, 4-bit8K-token context × 1 sequence | 66 GB | Fits | A100 80GB (80 GB) |
A job fits when the estimate is within 90% of the card's memory. Training assumes gradient checkpointing and FlashAttention; multi-GPU counts assume the memory splits evenly (FSDP, ZeRO-3 or tensor parallelism).
How we estimate it
Every figure comes from the model's published architecture (layers, hidden size, attention and KV heads, vocabulary) rather than a rule of thumb, using the same functions as the calculator above. Sizes are in GB (10⁹ bytes).
- Full fine-tuning: 16 bytes per parameter for mixed-precision Adam (bf16 weights and gradients, fp32 master weights and two moments), as in the ZeRO paper.
- LoRA: frozen bf16 weights (2 bytes per parameter) plus 16 bytes per adapter parameter, rank 16 on every linear layer. For mixture-of-experts models, adapters go on attention only and the experts stay frozen.
- QLoRA: NF4 weights with double quantisation (4.127 bits per parameter); embeddings and the output head stay in bf16, as bitsandbytes leaves them.
- Activations: gradient checkpointing and FlashAttention. Each layer's bf16 input is kept, one layer is recomputed at a time, and the output logits take 8 bytes per vocabulary entry per token (fp32 values and gradients). Chunked cross-entropy, as in Unsloth or Liger kernels, cuts that last term.
- Inference: 2, 1 or about 0.56 bytes per weight (4-bit with group-wise scales), plus an fp16 KV cache of 2 × layers × KV heads × head dim × 2 bytes per token; sliding-window layers only cache their window.
- Fits means the estimate is within 90% of the card's listed memory; the rest covers the CUDA context, kernels and fragmentation. Multi-GPU counts assume memory splits evenly.
Speed estimates use B200's dense BF16 peak of 2,250 TFLOPS, about 4 FLOPs per active parameter per token for LoRA and 40% utilisation; see fine-tuning and training costs for the method.
Frequently asked questions
- Can I fine-tune gpt-oss-120b on a B200?
- QLoRA: yes, about 66 GB at 2K tokens, batch 1, rank 16; up to 32K-token sequences at batch 1 or batch 24 at 2K, LoRA rank up to 1024. LoRA: no, it needs about 238 GB at 2K tokens, more than the 180 GB available. Full fine-tuning: no, it needs about 1,873 GB at 2K tokens, more than the 180 GB available.
- Can I run gpt-oss-120b on a B200?
- 16-bit (BF16/FP16) inference: no, the weights alone are 234 GB. 8-bit inference: yes, 117 GB of weights, room for 128K tokens of context or 290 concurrent 4K-token sequences. 4-bit inference: yes, 65 GB of weights, room for 128K tokens of context or 621 concurrent 4K-token sequences.
- What does fine-tuning gpt-oss-120b on a B200 cost?
- With QLoRA at our planning rate of 40% utilisation, about 5 minutes for 10 million training tokens, around $0.64 at $7.99 an hour (Runpod Secure Cloud, November 2026).
Related
- gpt-oss-120b VRAM requirementsEvery GPU and method
- gpt-oss-120b on RTX 509032 GB
- gpt-oss-120b on H100 SXM80 GB
- gpt-oss-120b on H200141 GB
- Llama 3.1 70B on B200Meta
Sources
- Hugging Face. openai/gpt-oss-120b config.json
- OpenAI. openai/gpt-oss-120b model card
- OpenAI. gpt-oss-120b & gpt-oss-20b Model Card (arXiv:2508.10925)
- NVIDIA. B200 specifications
- Runpod. GPU pricing (updated 12 November 2026)
- Lambda. GPU cloud pricing
- Dettmers et al., 2023. QLoRA: Efficient Finetuning of Quantized LLMs (arXiv:2305.14314)
- Hu et al., 2021. LoRA: Low-Rank Adaptation of Large Language Models (arXiv:2106.09685)
- Rajbhandari et al., 2019. ZeRO: Memory Optimizations Toward Training Trillion Parameter Models (arXiv:1910.02054)
- Korthikanti et al., 2022. Reducing Activation Recomputation in Large Transformer Models (arXiv:2205.05198)
Measure before you rent the GPU.
Prompt tuning and model search in Kojen run on hosted models against your own test cases, so you know whether fine-tuning gpt-oss-120b is worth the GPU time.
Start training free