Model tiers

How we group models — and how to override us.

Three quality tiers. When you opt in to cheaper-model swaps, we pick the cheapest model in the same tier as the one you asked for — not across tiers, not when tools are set. "Same tier" means published-benchmark equivalence within a small band, listed per tier below. The mapping is public, re-evaluated on every major model release, and overridable at the org or request level.

How we assign tiers

Each model gets a tier based on published benchmark scores (details per tier below). Assignments are hand-curated by us today — we don't run continuous evals ourselves. When a provider ships a new frontier model, we place it based on the benchmark numbers they publish and any early independent evals we trust; when the model has been out for a few weeks and community benchmarks stabilize, we re-check.

When a swap happens on a request you opted in to, the response carries ax-gateway-routed-fromheader so you can audit exactly which model answered. The arbitrage picker never crosses tiers, never swaps whentoolsis set on the request, and never swaps a vision-bearing request to a text-only model.

Every response also carries two audit headers you can verify independently:x-gateway-swap-reasonnames why we swapped (e.g., cheapest-in-tier), andx-gateway-benchmark-basislists the exact benchmarks that placed the served model in its tier. Cross-reference the benchmark list against the tier tables below — if we ever silently move a goalpost, the header disagrees with this page and you can call us on it.

Flagship

39 models

General-purpose top-line models. Long-form generation, complex reasoning, code, agentic multi-turn. What most product traffic runs on.

Benchmarks consulted: MMLU · GPQA · HumanEval · MT-Bench

Same-tier band: within ~3 points on MMLU + within ~5 on GPQA / HumanEval

Example swap: Claude Sonnet ↔ DeepSeek-V4-Flash — both flagship-band, ~10× price difference.

deepseek/deepseek-v4-1-flash

cheapest
Provider
deepseek
Region
global
Input /1M
$0.15
Output /1M
$0.60
Vision
Yes

groq/openai/gpt-oss-120b

Provider
groq
Region
global
Input /1M
$0.15
Output /1M
$0.60
Vision
No

sarvam/sarvam-105b

Provider
sarvam
Region
in
Input /1M
$0.16
Output /1M
$0.63
Vision
No

sarvam/sarvam-105b-conversations

Provider
sarvam
Region
in
Input /1M
$0.16
Output /1M
$0.63
Vision
No

qwen/qwen-mt-plus

Provider
qwen
Region
cn
Input /1M
$0.26
Output /1M
$0.78
Vision
No

glm/glm-4.6v

Provider
glm
Region
cn
Input /1M
$0.30
Output /1M
$0.90
Vision
Yes

glm/glm-4.5v

Provider
glm
Region
cn
Input /1M
$0.60
Output /1M
$1.80
Vision
Yes

doubao/doubao-seed-2.1-turbo

Provider
doubao
Region
cn
Input /1M
$0.42
Output /1M
$2.08
Vision
Yes

doubao/doubao-seed-2.0-pro

Provider
doubao
Region
cn
Input /1M
$0.44
Output /1M
$2.22
Vision
No

xai/grok-4-20-0309-non-reasoning

Provider
xai
Region
global
Input /1M
$1.25
Output /1M
$2.50
Vision
No

xai/grok-4-20-0309-reasoning

Provider
xai
Region
global
Input /1M
$1.25
Output /1M
$2.50
Vision
No

xai/grok-4-3

Provider
xai
Region
global
Input /1M
$1.25
Output /1M
$2.50
Vision
No

glm/glm-5

Provider
glm
Region
cn
Input /1M
$1.00
Output /1M
$3.20
Vision
No

kimi/kimi-k2.6

Provider
kimi
Region
cn
Input /1M
$0.95
Output /1M
$4.00
Vision
Yes

glm/glm-5v-turbo

Provider
glm
Region
cn
Input /1M
$1.20
Output /1M
$4.00
Vision
Yes

doubao/doubao-seed-2.1-pro

Provider
doubao
Region
cn
Input /1M
$0.83
Output /1M
$4.17
Vision
Yes

meta/muse-spark-1.1

Provider
meta
Region
global
Input /1M
$1.25
Output /1M
$4.25
Vision
No

meta/muse-spark-1.2

Provider
meta
Region
global
Input /1M
$1.25
Output /1M
$4.25
Vision
No

meta/muse-spark-1.3

Provider
meta
Region
global
Input /1M
$1.25
Output /1M
$4.25
Vision
No

glm/glm-5.1

Provider
glm
Region
cn
Input /1M
$1.40
Output /1M
$4.40
Vision
No

glm/glm-5.2

Provider
glm
Region
cn
Input /1M
$1.40
Output /1M
$4.40
Vision
No

glm/glm-5.3

Provider
glm
Region
cn
Input /1M
$1.40
Output /1M
$4.40
Vision
No

glm/glm-4.5-airx

Provider
glm
Region
cn
Input /1M
$1.10
Output /1M
$4.50
Vision
No

openai/gpt-6.1-terra

Provider
openai
Region
global
Input /1M
$1.00
Output /1M
$5.00
Vision
No

qwen/qwen3.8-max

Provider
qwen
Region
cn
Input /1M
$2.00
Output /1M
$6.00
Vision
No

xai/grok-4-5

Provider
xai
Region
global
Input /1M
$2.00
Output /1M
$6.00
Vision
No

xai/grok-4-6

Provider
xai
Region
global
Input /1M
$2.00
Output /1M
$6.00
Vision
No

xai/grok-4-7

Provider
xai
Region
global
Input /1M
$2.00
Output /1M
$6.00
Vision
No

qwen/qwen-max

Provider
qwen
Region
cn
Input /1M
$1.60
Output /1M
$6.40
Vision
No

qwen/qwen3.7-max

Provider
qwen
Region
cn
Input /1M
$2.50
Output /1M
$7.50
Vision
No

glm/glm-4.5-x

Provider
glm
Region
cn
Input /1M
$2.20
Output /1M
$8.90
Vision
No

anthropic/claude-sonnet-5

Provider
anthropic
Region
global
Input /1M
$2.00
Output /1M
$10.00
Vision
Yes

anthropic/claude-sonnet-5-5

Provider
anthropic
Region
global
Input /1M
$2.00
Output /1M
$10.00
Vision
Yes

google/gemini-4-argon

Provider
google
Region
global
Input /1M
$2.00
Output /1M
$10.00
Vision
Yes

kimi/kimi-k3.1-pro

Provider
kimi
Region
cn
Input /1M
$2.00
Output /1M
$10.00
Vision
No

openai/gpt-6.1-sol

Provider
openai
Region
global
Input /1M
$2.00
Output /1M
$10.00
Vision
No

openai/gpt-5.6-terra

Provider
openai
Region
global
Input /1M
$2.00
Output /1M
$12.00
Vision
Yes

anthropic/claude-sonnet-4-6

Provider
anthropic
Region
global
Input /1M
$3.00
Output /1M
$15.00
Vision
Yes

bedrock/anthropic.claude-sonnet-4-5

Provider
bedrock
Region
global
Input /1M
$3.00
Output /1M
$15.00
Vision
Yes

Cheap-Fast

31 models

Small, fast, cheap. Classification, short Q&A, structured extraction, tool-calling backbone for agents. Same shape as flagship traffic on the wire.

Benchmarks consulted: MMLU · HumanEval · BigBench-Hard

Same-tier band: within ~5 points on MMLU + HumanEval

Example swap: GPT-4o-mini ↔ Gemini 3.1 Flash Lite ↔ GPT-5 Nano — priced within roughly 2× of each other.

doubao/doubao-seed-2.0-mini

cheapest
Provider
doubao
Region
cn
Input /1M
$0.028
Output /1M
$0.28
Vision
No

qwen/qwen-vl-ocr

Provider
qwen
Region
cn
Input /1M
$0.043
Output /1M
$0.072
Vision
Yes

qwen/qwen-mt-flash

Provider
qwen
Region
cn
Input /1M
$0.10
Output /1M
$0.28
Vision
No

groq/openai/gpt-oss-20b

Provider
groq
Region
global
Input /1M
$0.075
Output /1M
$0.30
Vision
No

krutrim/openai/gpt-oss-20b

Provider
krutrim
Region
in
Input /1M
$0.084
Output /1M
$0.33
Vision
No

glm/glm-4.6v-flashx

Provider
glm
Region
cn
Input /1M
$0.040
Output /1M
$0.40
Vision
Yes

glm/glm-4.7-flashx

Provider
glm
Region
cn
Input /1M
$0.070
Output /1M
$0.40
Vision
No

qwen/qwen3.5-flash

Provider
qwen
Region
cn
Input /1M
$0.10
Output /1M
$0.40
Vision
No

qwen/qwen3.8-flash

Provider
qwen
Region
cn
Input /1M
$0.15
Output /1M
$0.47
Vision
No

doubao/doubao-seed-2.0-lite

Provider
doubao
Region
cn
Input /1M
$0.083
Output /1M
$0.50
Vision
No

anthropic/claude-haiku-5-5

Provider
anthropic
Region
global
Input /1M
$0.10
Output /1M
$0.50
Vision
Yes

openai/gpt-6-luna

Provider
openai
Region
global
Input /1M
$0.10
Output /1M
$0.50
Vision
No

glm/glm-5.3-flash

Provider
glm
Region
cn
Input /1M
$0.15
Output /1M
$0.50
Vision
No

krutrim/openai/gpt-oss-120b

Provider
krutrim
Region
in
Input /1M
$0.14
Output /1M
$0.68
Vision
No

glm/glm-4.5-air

Provider
glm
Region
cn
Input /1M
$0.20
Output /1M
$1.10
Vision
No

openai/gpt-5.6-luna

Provider
openai
Region
global
Input /1M
$0.20
Output /1M
$1.20
Vision
Yes

qwen/qwen-plus

Provider
qwen
Region
cn
Input /1M
$0.40
Output /1M
$1.20
Vision
No

glm/glm-5.3-flashx

Provider
glm
Region
cn
Input /1M
$0.37
Output /1M
$1.25
Vision
No

google/gemini-3.1-flash-lite

Provider
google
Region
global
Input /1M
$0.25
Output /1M
$1.50
Vision
Yes

qwen/qwen3.7-plus

Provider
qwen
Region
cn
Input /1M
$0.40
Output /1M
$1.60
Vision
No

xai/grok-build-0-1

Provider
xai
Region
global
Input /1M
$1.00
Output /1M
$2.00
Vision
No

glm/glm-4.5

Provider
glm
Region
cn
Input /1M
$0.60
Output /1M
$2.20
Vision
No

glm/glm-4.6

Provider
glm
Region
cn
Input /1M
$0.60
Output /1M
$2.20
Vision
No

glm/glm-4.7

Provider
glm
Region
cn
Input /1M
$0.60
Output /1M
$2.20
Vision
No

google/gemini-3.5-flash-lite

Provider
google
Region
global
Input /1M
$0.30
Output /1M
$2.50
Vision
Yes

google/gemini-3.6-flash

Provider
google
Region
global
Input /1M
$0.75
Output /1M
$3.75
Vision
Yes

google/gemini-3.8-flash

Provider
google
Region
global
Input /1M
$0.75
Output /1M
$3.75
Vision
Yes

kimi/kimi-k2.7-code

Provider
kimi
Region
cn
Input /1M
$0.95
Output /1M
$4.00
Vision
Yes

anthropic/claude-haiku-4-5

Provider
anthropic
Region
global
Input /1M
$1.00
Output /1M
$5.00
Vision
Yes

bedrock/anthropic.claude-haiku-4-5

Provider
bedrock
Region
global
Input /1M
$1.00
Output /1M
$5.00
Vision
Yes

kimi/kimi-k2.7-code-highspeed

Provider
kimi
Region
cn
Input /1M
$1.90
Output /1M
$8.00
Vision
Yes

Premium Reasoning

12 models

Deep step-by-step reasoning, math, hard code. What you route to when quality matters more than latency and cost.

Benchmarks consulted: GPQA · MATH-500 · LiveCodeBench · AIME

Same-tier band: within ~5 points on GPQA + MATH-500

Example swap: Claude Opus ↔ DeepSeek-Reasoner — same reasoning-tier band, ~10× cheaper.

deepseek/deepseek-v4-pro

cheapest
Provider
deepseek
Region
global
Input /1M
$0.66
Output /1M
$1.98
Vision
No

kimi/kimi-k3

Provider
kimi
Region
cn
Input /1M
$3.00
Output /1M
$15.00
Vision
No

anthropic/claude-opus-5-5

Provider
anthropic
Region
global
Input /1M
$4.00
Output /1M
$20.00
Vision
Yes

openai/gpt-5.6-sol

Provider
openai
Region
global
Input /1M
$4.00
Output /1M
$20.00
Vision
Yes

anthropic/claude-opus-4-5

Provider
anthropic
Region
global
Input /1M
$5.00
Output /1M
$25.00
Vision
Yes

anthropic/claude-opus-4-6

Provider
anthropic
Region
global
Input /1M
$5.00
Output /1M
$25.00
Vision
Yes

anthropic/claude-opus-4-7

Provider
anthropic
Region
global
Input /1M
$5.00
Output /1M
$25.00
Vision
Yes

anthropic/claude-opus-4-8

Provider
anthropic
Region
global
Input /1M
$5.00
Output /1M
$25.00
Vision
Yes

anthropic/claude-opus-5

Provider
anthropic
Region
global
Input /1M
$5.00
Output /1M
$25.00
Vision
Yes

anthropic/claude-fable-5

Provider
anthropic
Region
global
Input /1M
$10.00
Output /1M
$50.00
Vision
Yes

anthropic/claude-fable-5-1

Provider
anthropic
Region
global
Input /1M
$10.00
Output /1M
$50.00
Vision
Yes

openai/gpt-6-astra

Provider
openai
Region
global
Input /1M
$10.00
Output /1M
$50.00
Vision
No

Override the mapping.

Three levers. Use whichever fits your rollout.

Per-request: force the exact model

POST /v1/chat/completions
X-Gateway-Routing: explicit

With this header, the gateway sends to the exact model in your request'smodelfield, no swap. Useful for the small slice of traffic where you know model identity matters (deterministic user output, upstream tool-call quirks, benchmark runs). Applies per request; leave it off elsewhere to keep the savings.

Per-org: disable swaps entirely

Turn off cheaper-model swaps at /dashboard/settings → Routing. Every request routes to the exact model you asked for until you re-enable. Same effect as sending the header on every call, but org-wide.

Per-org: pin individual models

At /dashboard/settings → Tier overrides you can, per model, either move it to a different tier or exclude it from arbitrage entirely for your org. Common patterns: "treat Claude Haiku as flagship because we trust it", or "never let the picker route to DeepSeek-V4-Flash for us." Explicit requests still route to the requested model regardless of these settings.

What we don't do yet.

One honest gap worth naming so you can decide whether it matters to you.

  • No continuous eval. We hand-place new models based on the benchmark scores providers publish and community re-runs we trust. When a benchmark shifts materially or a model quietly degrades between versions, we'll notice on the next audit cycle, not the day it happens. If you care about that lag, keep swaps off org-wide, or use tier overrides to pin the specific models you care about, and re-visit as we build out nightly eval infrastructure.

See also: Savings Calculator · Full model catalog · Competitor comparisons