Imagine 1.0 or Imagine 2.0: Quality and Costs

Updated November 2, 2026. The “pro” model from the previous version of this course, grok-imagine-image-quality, was retired on November 2, 2026: its requests are now served by grok-imagine-image-2.0 in low quality. The current lineup includes two image models, presented here.

Choose the right generation model

xAI offers two image models. The first, Imagine 1.0, is the cheapest: it is the tool for exploration and volume. The second, Imagine 2.0, costs double but adds a quality setting, more formats, and editing from multiple images. The choice depends on what you do with the image: a draft among a hundred, or the visual that gets published.

The two available models

grok-imagine-image (Imagine 1.0)

• Cost: $0.02 per image
• Throughput: 6 requests per second (official model card)
• Ideal usage: Prototyping, prompt exploration, high-volume web content

response = client.images.generate(
    model="grok-imagine-image",
    prompt="Minimalist logo for a mobile fitness app",
)

grok-imagine-image-2.0 (Imagine 2.0)

• Cost: $0.04 per image in low quality; each image is billed at the quality at which it is served
• Quality setting: low, medium (more compute for more detail), or auto, the default value, which currently serves low in generation and medium in editing
• Plus: Up to five source images in editing, and 21:9 and 5:2 formats
• Throughput: 6 requests per second (official model card)
• Ideal usage: Final visuals, commercial illustrations, faithful edits

With the OpenAI SDK pointed at the xAI API, xAI-specific parameters, like quality, go through extra_body:

response = client.images.generate(
    model="grok-imagine-image-2.0",
    prompt="Hyperrealistic portrait of an architect in front of a futuristic skyscraper",
    extra_body={"quality": "medium"},
)

Comparison

Recommended usage strategy

Explore with what is cheap, finalize with what is good. A prompt takes ten or fifteen tries to refine; doing those with the most expensive model means paying final render rates for images you throw away.

1. Exploration phase: Imagine 1.0, at $0.02 per image, to test prompts and creative directions.
2. Finalization phase: Once the prompt is found, Imagine 2.0, in medium if detail matters more than cost.
3. Volume production: For a catalog or thumbnails, Imagine 1.0 remains the most economical; Imagine 2.0 in low if you need its formats or multi-image editing.

Cost estimation

For a project generating 1,000 images per month, at listed rates:

• Imagine 1.0 only: 1,000 × $0.02 = $20
• Imagine 2.0 in low only: 1,000 × $0.04 = $40
• Mixed (90% Imagine 1.0, 10% Imagine 2.0 in low): 900 × $0.02 + 100 × $0.04 = $22

The medium quality costs more: price it against the current rate card before applying it at volume.

Hands-on

Compare the two models on the same prompt:

prompt = "Mountain landscape at sunset, oil painting style"

v1 = client.images.generate(model="grok-imagine-image", prompt=prompt)
v2 = client.images.generate(
    model="grok-imagine-image-2.0",
    prompt=prompt,
    extra_body={"quality": "medium"},
)

print(f"Imagine 1.0: {v1.data[0].url}")
print(f"Imagine 2.0: {v2.data[0].url}")

Rates as of November 2, 2026 — prices change: before any budget calculation, check the current rate card on the xAI pricing page. Source: grok-imagine-image-quality Retirement on November 2, 2026.

Key takeaways

• Two models: Imagine 1.0 (grok-imagine-image, $0.02) and Imagine 2.0 (grok-imagine-image-2.0, $0.04 in low)
• grok-imagine-image-quality was retired on November 2, 2026: its requests route to Imagine 2.0 in low
• With the OpenAI SDK, quality, aspect_ratio, and resolution go through extra_body
• Explore with Imagine 1.0, finalize with Imagine 2.0