Top pick
GPT-5.6 Luna
$1.18 / month
Best cited image score for the bill: MMMU Pro 78.4 (openai.com).
Score 77.7 · $0.20 input · $1.20 output per 1M · context 1,050,000 · 46 providers
Pick for Image understanding: GPT-5.6 Luna at $1.18 per month. Best cited image score for the bill: MMMU Pro 78.4 (openai.com). Prices from Models.dev. Data last refreshed Oct 16, 2026, 07:15 AM UTC.
Job
Image understanding
Which model should I use to read images, and what will it cost each month?
The model has to accept images. The score uses the first cited bench in this order: MMMU Pro, then CharXiv Reasoning, then MathVision, then BabyVision. A narrower bench is capped so it does not outrank a multimodal score. Cards are chosen only from models with one of those cited results.
How this pick is chosen
- Outputs text. Embeddings and image-only models are not in the catalog.
- Lab is one of: openai, anthropic, google, meta, mistral, deepseek, xai, alibaba, moonshotai, zhipuai, cohere, amazon, nvidia, microsoft, ai21.
- Listed by at least 5 providers. A model on only the lab and a reseller does not clear this.
- Not deprecated, and not an alpha release. Beta releases stay in.
- Not a specialized model: OCR, embeddings, rerankers, speech, safeguard, or moderation, matched on the name.
- Accepts image input.
- Context window of at least 32,000 tokens.
- Has a numeric input and output price.
- Reference price is not $0. A zero in Models.dev is often an unpublished price.
- A benchmark counts only when its source is the model's lab (including that lab's Hugging Face org) or the benchmark owner. Artificial Analysis, OpenRouter, and LLM Stats are dropped. When one benchmark has several cited scores, the median is used.
- Image score: the first cited bench in this order, times a cap so a narrower bench cannot outrank a multimodal one: MMMU Pro (1), CharXiv Reasoning (0.85), MathVision (0.8), BabyVision (0.7). The monthly bill then subtracts 3 times its log. A model with none of those cited benches loses 30. The cards are chosen only from models with one of those cited scores.
- The top pick is the highest job score. On coding agents, models within 4 points of that leader are treated as a flagship tie: the lower monthly bill is the top pick and the higher bill is the premium pick. Otherwise the premium pick is the best upgrade that costs more than the top pick. The budget pick is the highest job score that costs less than the top pick and sits at or below the 25th cost percentile. Output rates above $50 per 1M tokens stay out of the premium card. Summarization uses raw quality, not quality per dollar, when it chooses that upgrade. Remaining ties break to more providers, then a lower monthly cost, then a newer release date.
Forty images a day. Each request counts about 2,500 input tokens, which is enough for a screenshot or a page photo plus a short question, and about 400 tokens of answer.
155 models cleared the bar.304 failed the lab, provider, or specialty bar.63 failed a capability filter.0 listed at $0 were left out.0 had no usable price.
Budget pick
Qwen3.8 Flash Next
$0.84 / month
Lower bill with a cited image score: CharXiv Reasoning 87.6 (huggingface.co).
Score 75.2 · $0.20 input · $0.50 output per 1M · context 262,144 · 5 providers
Premium pick
Gemini 3.5 Flash
$8.82 / month
Highest cited image score among the models that cost more: MMMU Pro 83.6 (deepmind.google).
Score 74.2 · $1.50 input · $9.00 output per 1M · context 1,048,576 · 34 providers
When not to use it
Skip this page for OCR-only tools, video, or a screenshot agent that has to click. Those are different jobs. A cited MMMU or chart score is not a promise about your own forms.
The reference workload is 40 images a day. Image tokens are priced as input tokens. The calculator changes the bill, not the rank.
The three cards follow the inputs. Fit, and who holds each role, stays on the reference workload.
Everyone who cleared the bar, best fit first
| Caps | ||||||||
|---|---|---|---|---|---|---|---|---|
| 77.7 | GPT-5.6 Luna | $1.18 | $0.20 | $1.20 | 1,050,000 | 46 | Jul 9, 2026 | |
| 75.2 | Qwen3.8 Flash Next | $0.84 | $0.20 | $0.50 | 262,144 | 5 | Aug 27, 2026 | |
| 74.2 | Gemini 3.5 Flash | $8.82 | $1.50 | $9.00 | 1,048,576 | 34 | May 19, 2026 | |
| 70.9 | GPT-5.4 mini | $4.41 | $0.75 | $4.50 | 400,000 | 33 | Mar 17, 2026 | |
| 70.0 | GPT-5.6 Terra | $11.76 | $2.00 | $12.00 | 1,050,000 | 43 | Jul 9, 2026 | |
| 69.7 | GPT-5.6 Sol | $21.60 | $4.00 | $20.00 | 1,050,000 | 47 | Jul 9, 2026 | |
| 69.6 | Qwen3.8 27B | $2.70 | $0.42 | $3.00 | 262,144 | 45 | Aug 14, 2026 | |
| 69.6 | GPT-5.4 | $14.70 | $2.50 | $15.00 | 1,050,000 | 40 | Mar 5, 2026 | |
| 68.2 | Gemini 3.6 Flash | $4.05 | $0.75 | $3.75 | 1,048,576 | 30 | Jul 21, 2026 | |
| 67.5 |  Muse Spark 1.1 | $5.79 | $1.25 | $4.25 | 1,048,576 | 15 | Apr 8, 2026 | |
| 67.2 | Gemini 3.8 Flash | $4.05 | $0.75 | $3.75 | 1,048,576 | 27 | Sep 2, 2026 | |
| 67.0 | GPT-5.4 nano | $1.20 | $0.20 | $1.25 | 400,000 | 29 | Mar 17, 2026 |