Skip to content

Add Grok 4.7 pricing and bill 200K-token xAI prompts at long rates #261

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mike1858 merged 3 commits into from
Sep 22, 2026
Merged

Add Grok 4.7 pricing and bill 200K-token xAI prompts at long rates #261

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1454e25
Add Grok 4.7 pricing
mike1858 Sep 22, 2026
8f55f22
Bill exactly 200K-token xAI prompts at long-context rates
mike1858 Sep 22, 2026
2600e84
Merge branch 'main' into models/grok-4-7-pricing
mike1858 Sep 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
128 changes: 114 additions & 14 deletions src/models.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2474,12 +2474,21 @@ fn populate_defaults(

// xAI Models
// Source: https://docs.x.ai/developers/pricing
//
// Every xAI tier boundary below is written as `Some(199_999)`, not
// `Some(200_000)`. xAI labels each long-context tier "≥ 200k tokens" and
// says the long rates apply once a prompt *reaches* the threshold, so a
// prompt of exactly 200,000 tokens must bill long. `find_tier` treats a
// limit as inclusive (`tokens <= limit` selects that tier), so the short
// tier has to stop one token earlier. Other vendors in this file publish
// "> 200K" boundaries and correctly keep `Some(200_000)`; do not unify the
// two spellings without re-reading each vendor's wording.
add_model!(
"grok-4.3",
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 1.25,
output_per_1m: 2.50,
},
Expand All @@ -2494,7 +2503,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.20,
},
CachingTier {
Expand All @@ -2511,7 +2520,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 1.25,
output_per_1m: 2.50,
},
Expand All @@ -2526,7 +2535,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.20,
},
CachingTier {
Expand All @@ -2543,7 +2552,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 1.25,
output_per_1m: 2.50,
},
Expand All @@ -2558,7 +2567,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.20,
},
CachingTier {
Expand All @@ -2575,7 +2584,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 1.25,
output_per_1m: 2.50,
},
Expand All @@ -2590,7 +2599,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.20,
},
CachingTier {
Expand All @@ -2607,7 +2616,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 2.00,
output_per_1m: 6.00,
},
Expand All @@ -2622,7 +2631,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.30,
},
CachingTier {
Expand All @@ -2639,7 +2648,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 2.00,
output_per_1m: 6.00,
},
Expand All @@ -2654,7 +2663,44 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.50,
},
CachingTier {
max_tokens: None,
cached_input_per_1m: 1.00,
},
],
bracket_pricing: true,
}),
false
);
// Grok 4.7 repeats Grok 4.6's schedule exactly: $2/$6 with $0.50 cached
// input below the 200K prompt boundary, doubling to $4/$12 with $1.00
// cached above it. xAI held the price flat across the capability bump
// rather than raising it, so the duplicated literals below are the real
// published rates and not a copy-paste that needs deduplicating.
add_model!(
"grok-4.7",
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(199_999),
input_per_1m: 2.00,
output_per_1m: 6.00,
},
PricingTier {
max_tokens: None,
input_per_1m: 4.00,
output_per_1m: 12.00,
},
],
bracket_pricing: true,
}),
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(199_999),
cached_input_per_1m: 0.50,
},
CachingTier {
Expand All @@ -2671,7 +2717,7 @@ fn populate_defaults(
PricingStructure::Tiered(TieredPricing {
tiers: vec![
PricingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
input_per_1m: 1.00,
output_per_1m: 2.00,
},
Expand All @@ -2686,7 +2732,7 @@ fn populate_defaults(
CachingSupport::Tiered(TieredCaching {
tiers: vec![
CachingTier {
max_tokens: Some(200_000),
max_tokens: Some(199_999),
cached_input_per_1m: 0.20,
},
CachingTier {
Expand Down Expand Up @@ -5431,6 +5477,11 @@ mod tests {
.expect("Grok 4.6 should exist")
.is_estimated
);
assert!(
!get_model_info("grok-4.7")
.expect("Grok 4.7 should exist")
.is_estimated
);
for model in [
"grok-4.3",
"grok-4.20-0309-reasoning",
Expand Down Expand Up @@ -5474,6 +5525,47 @@ mod tests {
),
17.0,
);
// Grok 4.7 must price identically to 4.6 on both sides of the 200K
// boundary. Asserting the same literals rather than comparing the two
// models keeps the test honest if xAI ever diverges them: a future
// 4.7-only price change fails here instead of silently agreeing with
// whatever 4.6 happens to be.
approx_eq(
calculate_total_cost_for_context_at(
"grok-4.7", 1_000_000, 1_000_000, 0, 1_000_000, 199_999, None,
),
8.5,
);
approx_eq(
calculate_total_cost_for_context_at(
"grok-4.7", 1_000_000, 1_000_000, 0, 1_000_000, 200_001, None,
),
17.0,
);
// Exactly 200,000 context tokens is the case the section comment is
// about: xAI's "≥ 200k" wording puts it in the long tier. The 199,999
// and 200,001 probes above both pass whether the boundary is inclusive
// or exclusive, so only this probe can tell the two apart. It runs for
// every tiered xAI model because they all share the one boundary rule,
// and a single model reverted to `Some(200_000)` would halve its
// exact-boundary bill without any other assertion noticing.
for (model, long_total) in [
("grok-4.3", 7.9),
("grok-4.20-0309-reasoning", 7.9),
("grok-4.20-0309-non-reasoning", 7.9),
("grok-4.20-multi-agent-0309", 7.9),
("grok-4.5", 16.6),
("grok-4.6", 17.0),
("grok-4.7", 17.0),
("grok-build-0.1", 6.4),
] {
approx_eq(
calculate_total_cost_for_context_at(
model, 1_000_000, 1_000_000, 0, 1_000_000, 200_000, None,
),
long_total,
);
}
approx_eq(
calculate_total_cost_for_context_at(
"grok-code-fast-1",
Expand Down Expand Up @@ -5502,6 +5594,14 @@ mod tests {
"grok-4.6", 1_000_000, 1_000_000, 0, 1_000_000, 2_000_000, None,
),
);
approx_eq(
calculate_total_cost_for_context_at(
"grok-4.7", 1_000_000, 1_000_000, 999_999, 1_000_000, 2_000_000, None,
),
calculate_total_cost_for_context_at(
"grok-4.7", 1_000_000, 1_000_000, 0, 1_000_000, 2_000_000, None,
),
);
}

#[test]
Expand Down
Loading