Pick for Long-document Q&A: Muse Spark 1.3 at $94.64 per month. Best mix of context window and the long-prompt price: 1,048,576-token window, billed at $1.25 per 1M input on this workload, MRCR 98.3. Prices from Models.dev. Data last refreshed Oct 14, 2026, 05:43 PM UTC.
Job
Which model should I use to ask questions of documents over 200,000 tokens, and what will it cost each month?
Models under 200,000 tokens of context are left out. The score weighs the context window, the input rate this long prompt pays, and a cited MRCR result from the lab. A long window is not a promise that the model uses the whole document well.
Fifteen long questions a day. Most of the bill is the document in the prompt, not the answer.
160 models cleared the bar.303 failed the lab, provider, or specialty bar.56 failed a capability filter.0 listed at $0 were left out.0 had no usable price.
Top pick
Muse Spark 1.3$94.64 / month
Best mix of context window and the long-prompt price: 1,048,576-token window, billed at $1.25 per 1M input on this workload, MRCR 98.3.
Score 111.9 · $1.25 input · $4.25 output per 1M · context 1,048,576 · 16 providers
Budget pick
DeepSeek V4 Flash$10.60 / month
Best long-context value among the cheaper models: 1,000,000-token window, billed at $0.14 per 1M input on this workload, MRCR 78.7.
Score 94.7 · $0.14 input · $0.28 output per 1M · context 1,000,000 · 52 providers
Premium pick
Qwen3.8 Max$151 / month
Best long-context mix among the models that cost more: 1,000,000-token window, billed at $2.00 per 1M input on this workload, MRCR 92.9.
Score 106.1 · $2.00 input · $6.00 output per 1M · context 1,000,000 · 32 providers
Skip this pick when the documents are short enough that a smaller window is cheaper, when you need retrieval over a corpus instead of one prompt, or when you want a lab-published long-context result and the card does not cite one. MRCR is one needle test, not a full reading of your files.
The reference workload is fifteen questions a day, about 180,000 input tokens and 1,200 output tokens, with a 10% cache hit. Most of the bill is the document. Prompt caching changes the number more here than on short chats.
The three cards follow the inputs. Fit, and who holds each role, stays on the reference workload.
| Caps | ||||||||
|---|---|---|---|---|---|---|---|---|
| 111.9 | Muse Spark 1.3 | $94.64 | $1.25 | $4.25 | 1,048,576 | 16 | Sep 2, 2026 | |
| 107.6 | DeepSeek V4 Pro | $98.71 | $1.32 | $3.96 | 1,000,000 | 58 | Apr 24, 2026 | |
| 106.1 | Qwen3.8 Max | $151 | $2.00 | $6.00 | 1,000,000 | 32 | Aug 3, 2026 | |
| 102.9 | Muse Spark 1.2 | $94.64 | $1.25 | $4.25 | 1,048,576 | 18 | Aug 5, 2026 | |
| 94.8 | Qwen3.7 Plus | $30.35 | $0.40 | $1.60 | 1,000,000 | 30 | Jun 2, 2026 | |
| 94.8 | Qwen3.5 Plus | $33.70 | $0.40 | $2.40 | 1,000,000 | 11 | Feb 16, 2026 | |
| 94.7 | DeepSeek V4 Flash | $10.60 | $0.14 | $0.28 | 1,000,000 | 52 | Apr 24, 2026 | |
| 93.8 |  Nemotron 3 Ultra 550B A55B | $39.02 | $0.50 | $2.50 | 1,000,000 | 22 | Jun 4, 2026 | |
| 93.8 | Qwen3.6 Plus | $38.48 | $0.50 | $3.00 | 1,000,000 | 21 | Apr 2, 2026 | |
| 92.6 | GPT-6 Luna | $7.64 | $0.10 | $0.50 | 1,050,000 | 32 | Sep 22, 2026 | |
| 92.5 |  Mistral Large 4 | $51.27 | $0.68 | $2.09 | 1,048,576 | 9 | Oct 6, 2026 | |
| 91.4 | GPT-5.6 Luna | $15.39 | $0.20 | $1.20 | 1,050,000 | 44 | Jul 9, 2026 |
