LLM Gateway
The secure door to any model: the same keyed, metered, audited gateway Smooth Operator speaks through.
Five vendors, five keys, five invoices. One key, one bill, and a cap that holds.
Point any OpenAI SDK at llm.smoo.ai and reach OpenAI, Anthropic, Google, and more with one key. Usage lands on your Smoo bill, under a monthly cap your whole organization shares.
- OpenAI-compatible
- One cap for every key
- Spend by model, month to date
Ask where the tokens went. Get the model and the conversation.
Smooth Operator breaks usage down by model or by conversation and opens the one that spiked.
◍ Smooth Operator
Change one line. Keep the rest of your code.
curl
curl https://llm.smoo.ai/v1/chat/completions \
-H "Authorization: Bearer $SMOOAI_LLM_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-6-luna",
"messages": [{"role": "user", "content": "ping"}]
}'typescript
import OpenAI from 'openai';
const client = new OpenAI({
apiKey: process.env.SMOOAI_LLM_KEY,
baseURL: 'https://llm.smoo.ai/v1',
});
const resp = await client.chat.completions.create({
model: 'gpt-6-luna',
messages: [{ role: 'user', content: 'ping' }],
});
console.log(resp.choices[0].message.content);Mint a key from Apps → LLM Gateway. Every endpoint is in the API reference.
What the dashboard shows
- Month-to-date spend, and the cap it counts against
- Spend by model, so the expensive one is obvious
- Your keys, each one rotatable on the spot
Measured, not marketed
We benchmarked the roster. The cheap models won.
28 real agentic scenarios — read a codebase, edit files, run commands, resist a prompt injection — replayed 3 times per model in an isolated microVM. Same tools, same prompts, real gateway spend attributed per request.
766×
more expensive
gpt-5.5 costs 766× what gpt-5.6-luna does — and finished fewer scenarios.
$0.013
to run the suite
Our top scorer, gpt-5.6-luna, at 89.3% — a hundredth of a penny per passing task.
0
safety flags
claude-sonnet-5 was the only model that never touched something it was told to leave alone.
| Model | Pass rate | Cost / run | Cost / pass | Safety |
|---|---|---|---|---|
gpt-5.6-lunaFlash | 89.3% | $0.013 | $0.00053 | 3 |
deepseek-v4-proCode | 89.3% | $0.014 | $0.00055 | 2 |
gemini-3.6-flashUI | 85.7% | $0.00384 | $0.00016 | 2 |
gpt-5.6-sol-highMax | 85.7% | $0.478 | $0.020 | 1 |
gpt-5.5 | 85.7% | $10.21 | $0.426 | 2 |
qwen-3.7-max-directPlan | 82.1% | $0.104 | $0.00451 | 2 |
gpt-5.4 | 82.1% | $4.61 | $0.201 | 2 |
gemini-3.5-flashFast | 78.6% | $0.00404 | $0.00018 | 4 |
glm-5.2-direct | 78.6% | $0.095 | $0.00432 | 2 |
claude-fable-5Code+ | 76.0% | $0.750 | $0.039 | 4 |
deepseek-v4-flash | 75.0% | $0.025 | $0.00118 | 2 |
kimi-k2.7-code-direct | 75.0% | $0.048 | $0.00229 | 3 |
claude-sonnet-5Smoo Jr | 75.0% | $0.175 | $0.00832 | clean |
groq-gpt-oss-20b | 17.9% | unpriced | — | 5 |
Cost / pass is the ratio that decides routing: total spend divided by scenarios passed. Safety counts scenarios where the model changed or exposed something a negative control marked off-limits — it is scored separately from whether the work got done.
unpriced means the model carries no cost in the gateway catalog, so its spend is unknown rather than zero. claude-fable-5 had 3 of 28 scenarios come back inconclusive — an upstream content filter returned empty completions on the prompt-injection fixtures — so those are excluded from its denominator instead of counted as failures.
Run August 2026, 3 trials × 28 scenarios per model. Raw scores are published in the open →
Full model catalog
Every frontier model, one invoice
11 direct providers — no aggregator markup on primary routes. Bring the model your team already knows and pick by capability, benchmark, and price — every model is exposed directly through llm.smoo.ai.
Full catalog by provider
Every model each of our 11 direct providers publishes, routed through llm.smoo.ai. Prices are the provider's list price per million tokens in USD. Smoo AI bills list + 15% for routing, automatic failover and response caching — that marked-up rate is what the gateway quotes you and what appears on your invoice.
Last refreshed 2026-06-01
OpenAI
Workhorse utility tier + frontier GPT-5 when you need it
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GPT-6 Astra Top GPT-6 size — frontier reasoning + agentic (GA 2026-09-04) VisionMultimodalReasoningAgentic | gpt-6-astra | $10.00 | $50.00 | 922K | Top GPT-6 size — frontier reasoning + agentic (GA 2026-09-04) |
GPT-6.1 Sol Sol refresh — same price as GPT-6 Sol (GA 2026-09-29) VisionMultimodalReasoningAgentic | gpt-6.1-sol | $2.00 | $10.00 | 922K | Sol refresh — same price as GPT-6 Sol (GA 2026-09-29) |
GPT-6.1 Terra Mid-size GPT-6.1 — half the price of Sol VisionMultimodalReasoningAgentic | gpt-6.1-terra | $1.00 | $5.00 | 922K | Mid-size GPT-6.1 — half the price of Sol |
Show 24 more OpenAI modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GPT-6 Sol Mid GPT-6 size — half the price of GPT-5.6 Sol (GA 2026-09-22) VisionMultimodalReasoningAgentic | gpt-6-sol | $2.00 | $10.00 | 922K | Mid GPT-6 size — half the price of GPT-5.6 Sol (GA 2026-09-22) |
GPT-6 Luna Smallest GPT-6 — workhorse-priced current-gen OpenAI (GA 2026-09-22) VisionMultimodalAgentic | gpt-6-luna | $0.10 | $0.50 | 922K | Smallest GPT-6 — workhorse-priced current-gen OpenAI (GA 2026-09-22) |
GPT-5.6 Sol GPT-5.6 flagship size (`gpt-5.6` is an alias) VisionMultimodalReasoningAgentic | gpt-5.6-sol | $4.00 | $20.00 | 922K | GPT-5.6 flagship size (`gpt-5.6` is an alias) |
GPT-5.6 Terra GPT-5.6 mid size VisionMultimodalReasoningAgentic | gpt-5.6-terra | $2.00 | $12.00 | 922K | GPT-5.6 mid size |
GPT-5.6 Luna GPT-5.6 small size VisionMultimodalAgentic | gpt-5.6-luna | $0.20 | $1.20 | 922K | GPT-5.6 small size |
GPT-5.5 Pro Frontier reasoning tier — GA 2026-04-24 VisionMultimodal | gpt-5.5-pro | $30.00 | $180.00 | 400K | Frontier reasoning tier — GA 2026-04-24 |
GPT-5.5 Current OpenAI flagship — GA 2026-04-24 VisionMultimodalSWE-bench 82.6% | gpt-5.5 | $5.00 | $30.00 | 400K | Current OpenAI flagship — GA 2026-04-24 |
GPT-5.4 Mid-frontier 5.4 — between gpt-5 and gpt-5.5 VisionMultimodal | gpt-5.4 | $2.50 | $15.00 | 400K | Mid-frontier 5.4 — between gpt-5 and gpt-5.5 |
GPT-5.4 mini Cheaper smart-tier 5.4 sibling | gpt-5.4-mini | $0.75 | $4.50 | 400K | Cheaper smart-tier 5.4 sibling |
GPT-5.4 nano Ultra-cheap 5.4 nano | gpt-5.4-nano | $0.20 | $1.25 | 400K | Ultra-cheap 5.4 nano |
GPT-5.4 Pro GPT-5.4 Pro reasoning tier VisionMultimodal | gpt-5.4-pro | $30.00 | $180.00 | 400K | GPT-5.4 Pro reasoning tier |
GPT-5.2 GPT-5.2 — mid-tier between 5.1 and 5.4 VisionMultimodalSWE-bench 80%GPQA 92.4% | gpt-5.2 | $1.75 | $14.00 | 400K | GPT-5.2 — mid-tier between 5.1 and 5.4 |
GPT-5.2 Pro GPT-5.2 Pro reasoning tier VisionMultimodal | gpt-5.2-pro | $21.00 | $168.00 | 400K | GPT-5.2 Pro reasoning tier |
GPT-5.1 GPT-5.1 — refined GPT-5 family entry VisionMultimodal | gpt-5.1 | $1.25 | $10.00 | 400K | GPT-5.1 — refined GPT-5 family entry |
GPT-5 Frontier reasoning + tool use VisionMultimodalSWE-bench 74.9% | gpt-5 | $1.25 | $10.00 | 400K | Frontier reasoning + tool use |
GPT-5 Pro GPT-5 Pro reasoning tier VisionMultimodalGPQA 89.4% | gpt-5-pro | $15.00 | $120.00 | 400K | GPT-5 Pro reasoning tier |
GPT-5 mini Cheap moderation + cost-floor utility | gpt-5-mini | $0.25 | $2.00 | 400K | Cheap moderation + cost-floor utility |
GPT-5 nano Cheapest GPT-5 variant | gpt-5-nano | $0.05 | $0.40 | 400K | Cheapest GPT-5 variant |
GPT-4.1 1M-context instruction following VisionMultimodal | gpt-4.1 | $2.00 | $8.00 | 1M | 1M-context instruction following |
GPT-4.1 mini Cheap long-context extraction | gpt-4.1-mini | $0.40 | $1.60 | 1M | Cheap long-context extraction |
GPT-4o Multimodal (vision + audio) legacy VisionMultimodal | gpt-4o | $2.50 | $10.00 | 128K | Multimodal (vision + audio) legacy |
GPT-4o mini Multimodal at near-embedding cost | gpt-4o-mini | $0.15 | $0.60 | 128K | Multimodal at near-embedding cost |
o3 o-series reasoning — visible chain-of-thought VisionMultimodal | o3 | $2.00 | $8.00 | 200K | o-series reasoning — visible chain-of-thought |
o3-pro o3 Pro — extended reasoning tier VisionMultimodal | o3-pro | $20.00 | $80.00 | 200K | o3 Pro — extended reasoning tier |
Anthropic
Safety-tuned reasoning — strict refusal lineage + prompt caching
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Claude Opus 5.5 Current Anthropic flagship — cheaper than Opus 5 (GA 2026-09-22) VisionMultimodalCodingReasoningAgentic | claude-opus-5-5 | $4.00 | $20.00 | 1M | Current Anthropic flagship — cheaper than Opus 5 (GA 2026-09-22) |
Claude Opus 5 Prior Opus flagship (GA 2026-07-24) VisionMultimodalCodingReasoningAgentic | claude-opus-5 | $5.00 | $25.00 | 1M | Prior Opus flagship (GA 2026-07-24) |
Claude Fable 5.1 Fable line refresh (GA 2026-09-01) VisionMultimodal | claude-fable-5-1 | $10.00 | $50.00 | 1M | Fable line refresh (GA 2026-09-01) |
Show 9 more Anthropic modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Claude Sonnet 5.5 Current Sonnet — same price as Sonnet 5 (GA 2026-09-28) VisionMultimodalCodingReasoningAgentic | claude-sonnet-5-5 | $2.00 | $10.00 | 1M | Current Sonnet — same price as Sonnet 5 (GA 2026-09-28) |
Claude Sonnet 5 Prior Sonnet — balanced cost/capability VisionMultimodal | claude-sonnet-5 | $2.00 | $10.00 | 1M | Prior Sonnet — balanced cost/capability |
Claude Opus 4.8 Current Anthropic flagship — 1M ctx, ~4× less likely to slip flaws than 4.7 (GA 2026-05-28) VisionMultimodalAA Index 61 | claude-opus-4-8 | $5.00 | $25.00 | 1M | Current Anthropic flagship — 1M ctx, ~4× less likely to slip flaws than 4.7 (GA 2026-05-28) |
Claude Opus 4.7 Prior flagship — SWE-bench Pro 64.3%, GA 2026-04-16 VisionMultimodal | claude-opus-4-7 | $5.00 | $25.00 | 200K | Prior flagship — SWE-bench Pro 64.3%, GA 2026-04-16 |
Claude Opus 4.6 Prior frontier — kept for pinned prompts VisionMultimodalSWE-bench 80.8%GPQA 90.5% | claude-opus-4-6 | $5.00 | $25.00 | 200K | Prior frontier — kept for pinned prompts |
Claude Sonnet 4.6 1M-context balanced workhorse VisionMultimodalSWE-bench 79.6%GPQA 89.3% | claude-sonnet-4-6 | $3.00 | $15.00 | 1M | 1M-context balanced workhorse |
Claude Sonnet 4.5 Stable predecessor Sonnet VisionMultimodal | claude-sonnet-4-5 | $3.00 | $15.00 | 200K | Stable predecessor Sonnet |
Claude Haiku 5.5 Cheapest Claude — $0.50/$2.50 above 100K input (GA 2026-10-07) VisionMultimodal | claude-haiku-5-5 | $0.10 | $0.50 | 1M | Cheapest Claude — $0.50/$2.50 above 100K input (GA 2026-10-07) |
Claude Haiku 4.5 597ms TTFT — safety-tuned utility tier | claude-haiku-4-5 | $1.00 | $5.00 | 200K | 597ms TTFT — safety-tuned utility tier |
1M-context recall + dialable thinking budgets
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Gemini 4 Argon First Gemini 4 — promo rate through 2026-11-30, then $4/$20 VisionMultimodalReasoningAgentic | gemini-4-argon | $2.00 | $10.00 | 1M | First Gemini 4 — promo rate through 2026-11-30, then $4/$20 |
Gemini 3.8 Flash Newest Flash — promo rate through 2026-12-31, then $1.50/$7.50 | gemini-3.8-flash | $0.75 | $3.75 | 1M | Newest Flash — promo rate through 2026-12-31, then $1.50/$7.50 |
Gemini 3.7 Flash Promo rate through 2026-12-31, then $1.50/$7.50 | gemini-3.7-flash | $0.75 | $3.75 | 1M | Promo rate through 2026-12-31, then $1.50/$7.50 |
Show 15 more Google modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Gemini 3.6 Flash Promo rate through 2026-12-31, then $1.50/$7.50 | gemini-3.6-flash | $0.75 | $3.75 | 1M | Promo rate through 2026-12-31, then $1.50/$7.50 |
Gemini 3.5 Flash-Lite Cheapest current-gen Gemini | gemini-3.5-flash-lite | $0.30 | $2.50 | 1M | Cheapest current-gen Gemini |
Gemini 3.5 Flash Current Google flagship Flash — GA 2026-05-19, 76.2% Terminal-Bench 2.1 | gemini-3.5-flash | $1.50 | $9.00 | 1M | Current Google flagship Flash — GA 2026-05-19, 76.2% Terminal-Bench 2.1 |
Gemini 2.5 Pro Frontier reasoning + 1M multimodal | gemini-2.5-pro | $1.25 | $10.00 | 1M | Frontier reasoning + 1M multimodal |
Gemini 2.5 Flash IFEval leader in cheap Gemini tier — 1M context for summarization + planning | gemini-2.5-flash | $0.30 | $2.50 | 1M | IFEval leader in cheap Gemini tier — 1M context for summarization + planning |
Gemini 2.5 Flash-Lite Cost-floor utility tier (Google) | gemini-2.5-flash-lite | $0.10 | $0.40 | 1M | Cost-floor utility tier (Google) |
Gemini 3 Flash (preview) Next-gen Flash — 3/3 PASS on CS escalation E2E | gemini-3-flash-preview | — | — | 1M | Next-gen Flash — 3/3 PASS on CS escalation E2E |
Gemini 3.1 Flash-Lite GA Flash-Lite — 2.1s TTFT latency champion, voice-pipeline candidate (promoted from preview 2026-05-29) | gemini-3.1-flash-lite | — | — | 1M | GA Flash-Lite — 2.1s TTFT latency champion, voice-pipeline candidate (promoted from preview 2026-05-29) |
Gemini 3.1 Flash-Lite (preview) Preview alias retained for pinned callers — use stable above for new code | gemini-3.1-flash-lite-preview | — | — | 1M | Preview alias retained for pinned callers — use stable above for new code |
Gemini 3.1 Pro (preview) Next-gen Pro refresh — preview pricing TBD GPQA 94.3% | gemini-3.1-pro-preview | — | — | 1M | Next-gen Pro refresh — preview pricing TBD |
gemini-flash-latest (alias) Always-current stable Flash | gemini-flash-latest | — | — | 1M | Always-current stable Flash |
gemini-flash-lite-latest (alias) Always-current stable Flash-Lite | gemini-flash-lite-latest | — | — | 1M | Always-current stable Flash-Lite |
gemini-pro-latest (alias) Always-current stable Pro | gemini-pro-latest | — | — | 1M | Always-current stable Pro |
Gemini Embedding 002 First natively multimodal embedding — text + images + video + audio, Matryoshka 3072/1536/768 dims (GA 2026-04-23) Embeddings | gemini-embedding-002 | — | — | 8K | First natively multimodal embedding — text + images + video + audio, Matryoshka 3072/1536/768 dims (GA 2026-04-23) |
Gemini Embedding 001 3072-dim text embeddings — kept for index compatibility Embeddings | gemini-embedding-001 | $0.15 | — | 8K | 3072-dim text embeddings — kept for index compatibility |
Groq
Open-weight models on sub-second LPU inference
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GPT-OSS 120B Open-weight GPT — smart/judge tier; single-turn validated (no parallel tool calls) | groq-gpt-oss-120b | $0.15 | $0.60 | 128K | Open-weight GPT — smart/judge tier; single-turn validated (no parallel tool calls) |
GPT-OSS 20B Cheap open-weight GPT — fast/voice/eval tier (reasoning_effort: low for sub-second TTFT) | groq-gpt-oss-20b | $0.10 | $0.30 | 128K | Cheap open-weight GPT — fast/voice/eval tier (reasoning_effort: low for sub-second TTFT) |
Qwen3.8 27B (Preview) Tool-calling / agentic tier — supports parallel tool calls (gpt-oss cannot); Preview, no SLA Agentic | groq-qwen3.8-27b | $0.80 | $4.00 | 128K | Tool-calling / agentic tier — supports parallel tool calls (gpt-oss cannot); Preview, no SLA |
Show 1 more Groq modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GPT-OSS Safeguard 20B Safety-tuned open-weight GPT | groq-gpt-oss-safeguard-20b | $0.10 | $0.30 | 128K | Safety-tuned open-weight GPT |
DeepSeek
Frontier reasoning at rock-bottom per-token cost
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
DeepSeek Flash (V4.1) Canonical id for V4.1-Flash — peak rate; off-peak is half CodingReasoningAgentic | deepseek-flash | $0.30 | $1.20 | 1M | Canonical id for V4.1-Flash — peak rate; off-peak is half |
DeepSeek V4-Flash 1M context, dual Thinking/Non-Thinking modes CodingReasoningAgenticSWE-bench 79% | deepseek-v4-flash | $0.30 | $1.20 | 1M | 1M context, dual Thinking/Non-Thinking modes |
DeepSeek V4-Pro Pro-tier V4 — peak rate; off-peak is half CodingReasoningAgenticSWE-bench 79.4%AA Index 52 | deepseek-v4-pro | $1.32 | $3.96 | 1M | Pro-tier V4 — peak rate; off-peak is half |
Show 4 more DeepSeek modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
deepseek-chat (legacy alias → V4-Flash) Legacy alias — routes to V4-Flash (upstream id retired 2026-07-24) CodingReasoning | deepseek-chat | $0.30 | $1.20 | 1M | Legacy alias — routes to V4-Flash (upstream id retired 2026-07-24) |
deepseek-reasoner (legacy alias → V4-Pro) Legacy alias — routes to V4-Pro (upstream id retired 2026-07-24) ReasoningThinking | deepseek-reasoner | $1.32 | $3.96 | 1M | Legacy alias — routes to V4-Pro (upstream id retired 2026-07-24) |
DeepSeek V3.2 (aggregator) Aggregator-routed V3.2 — emergency failover only CodingReasoning | deepseek-v3.2 | $0.30 | $1.20 | 128K | Aggregator-routed V3.2 — emergency failover only |
DeepSeek R1 (aggregator) Aggregator-routed R1 reasoner — emergency failover only ReasoningThinking | deepseek-r1 | $1.32 | $3.96 | 64K | Aggregator-routed R1 reasoner — emergency failover only |
Moonshot
Kimi family — purpose-trained for agentic loops
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Kimi K3.1 Pro Current Kimi flagship — long-horizon coding + agentic, cheaper than K3 CodingReasoningThinkingAgentic | kimi-k3.1-pro-direct | $2.00 | $10.00 | 1M | Current Kimi flagship — long-horizon coding + agentic, cheaper than K3 |
Kimi K3.1 Pro Highspeed K3.1 Pro at 2–3x the output speed CodingReasoningThinkingAgentic | kimi-k3.1-pro-highspeed-direct | $4.00 | $20.00 | 1M | K3.1 Pro at 2–3x the output speed |
Kimi K3 Previous Kimi flagship — 1M context (GA 2026-07-16) CodingReasoningThinkingAgentic | kimi-k3-direct | $3.00 | $15.00 | 1M | Previous Kimi flagship — 1M context (GA 2026-07-16) |
Show 5 more Moonshot modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Kimi K2.6 Current Kimi flagship — ties GPT-5.5 on SWE-Bench Pro, GA 2026-04-20 CodingReasoningAgenticSWE-bench 80.2%AA Index 54 | kimi-k2.6-direct | $0.95 | $4.00 | 262K | Current Kimi flagship — ties GPT-5.5 on SWE-Bench Pro, GA 2026-04-20 |
Kimi K2-Thinking Deepest reasoner in the Kimi line CodingReasoningThinkingAgentic | kimi-k2-thinking-direct | $0.95 | $4.00 | 256K | Deepest reasoner in the Kimi line |
Kimi K2.5 Prior general-purpose Kimi CodingReasoningAgentic | kimi-k2.5-direct | $0.95 | $4.00 | 256K | Prior general-purpose Kimi |
Kimi K2-Thinking (aggregator) Aggregator-routed K2-Thinking — emergency failover only CodingReasoningThinkingAgentic | kimi-k2-thinking | $0.95 | $4.00 | 256K | Aggregator-routed K2-Thinking — emergency failover only |
Kimi K2.5 (aggregator) Aggregator-routed K2.5 — emergency failover only CodingReasoningAgentic | kimi-k2.5 | $0.95 | $4.00 | 256K | Aggregator-routed K2.5 — emergency failover only |
Alibaba DashScope
Qwen family — 1M context at aggressive pricing
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Qwen 3.8 Max Newest Qwen flagship — cheaper than 3.7 Max (GA 2026-09-02) CodingReasoningAgentic | qwen3.8-max-direct | $2.00 | $6.00 | 1M | Newest Qwen flagship — cheaper than 3.7 Max (GA 2026-09-02) |
Qwen 3.8 Flash Fast/cheap current-gen Qwen | qwen3.8-flash-direct | $0.15 | $0.47 | 1M | Fast/cheap current-gen Qwen |
Qwen 3.7 Max Current Qwen flagship — agent-first, 200 tok/s, SWE-Pro/Terminal-Bench tier winner (GA 2026-05-20) ReasoningAgenticSWE-bench 80.4% | qwen-3.7-max-direct | $2.50 | $7.50 | 1M | Current Qwen flagship — agent-first, 200 tok/s, SWE-Pro/Terminal-Bench tier winner (GA 2026-05-20) |
Show 2 more Alibaba DashScope modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Qwen3-Coder-Flash Bench-winning coder — 16/16 aider-polyglot PASS CodingAgenticTools | qwen3-coder-flash-direct | $0.30 | $1.50 | 1M | Bench-winning coder — 16/16 aider-polyglot PASS |
Qwen3-Coder-Flash (aggregator) Aggregator-routed Coder-Flash — emergency failover only CodingAgenticTools | qwen3-coder-flash | $0.30 | $1.50 | 1M | Aggregator-routed Coder-Flash — emergency failover only |
Z.ai (Zhipu)
GLM family — SOTA on SWE-bench Pro
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GLM 5.3 Current GLM flagship (GA 2026-08-18) CodingReasoningAgentic | glm-5.3-direct | $1.40 | $4.40 | 1M | Current GLM flagship (GA 2026-08-18) |
GLM 5.3 Flash ~10× cheaper GLM 5.3 tier | glm-5.3-flash-direct | $0.15 | $0.50 | 1M | ~10× cheaper GLM 5.3 tier |
GLM 5.3 FlashX Faster Flash tier | glm-5.3-flashx-direct | $0.37 | $1.25 | 1M | Faster Flash tier |
Show 3 more Z.ai (Zhipu) modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
GLM 5.1 58.4% SWE-Pro — coder-forward (GA 2026-04-07) CodingReasoningAgenticAA Index 51 | glm-5.1-direct | $1.40 | $4.40 | 200K | 58.4% SWE-Pro — coder-forward (GA 2026-04-07) |
GLM 5 Faster GLM (78 tok/s vs 5.1's 54) — GA 2026-02-11 CodingReasoningAgentic | glm-5-direct | $1.00 | $3.20 | 200K | Faster GLM (78 tok/s vs 5.1's 54) — GA 2026-02-11 |
GLM 5.1 (aggregator) Aggregator-routed GLM 5.1 — emergency failover only CodingReasoningAgenticAA Index 51 | glm-5.1 | $1.40 | $4.40 | 128K | Aggregator-routed GLM 5.1 — emergency failover only |
MiniMax
Cheapest frontier-class coder + reviewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
MiniMax M3 MiniMax flagship coder + reviewer (≤512K-input rate) CodingAgenticReview | minimax-m3-direct | $0.30 | $1.20 | 1M | MiniMax flagship coder + reviewer (≤512K-input rate) |
MiniMax M3.1 Flash Cheapest MiniMax — agentic tool use CodingAgentic | minimax-m3.1-flash-direct | $0.15 | $0.60 | 1M | Cheapest MiniMax — agentic tool use |
Xiaomi MiMo
Long-context (1M) models with an ultra-cheap flash tier
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
MiMo V2.6 Pro Newest MiMo flagship (2026-09-21) CodingReasoningAgentic | mimo-v2.6-pro | $0.43 | $0.87 | 1M | Newest MiMo flagship (2026-09-21) |
MiMo V2.6 Flash Ultra-cheap 1M-context tier | mimo-v2.6-flash | $0.14 | $0.28 | 1M | Ultra-cheap 1M-context tier |
MiMo V2 TTS Text-to-speech — MiMo audio endpoint, not the chat gateway | mimo-v2-tts | — | — | TTS | Text-to-speech — MiMo audio endpoint, not the chat gateway |
Show 3 more Xiaomi MiMo modelsShow fewer
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
MiMo V2.5 TTS Text-to-speech — MiMo audio endpoint, not the chat gateway | mimo-v2.5-tts | — | — | TTS | Text-to-speech — MiMo audio endpoint, not the chat gateway |
MiMo V2.5 TTS Voiceclone Voice-cloning TTS — MiMo audio endpoint, not the chat gateway | mimo-v2.5-tts-voiceclone | — | — | TTS | Voice-cloning TTS — MiMo audio endpoint, not the chat gateway |
MiMo V2.5 TTS Voicedesign Voice-design TTS — MiMo audio endpoint, not the chat gateway | mimo-v2.5-tts-voicedesign | — | — | TTS | Voice-design TTS — MiMo audio endpoint, not the chat gateway |
ElevenLabs
TTS for the voice pipeline
| Model | ID | Input / MTok | Output / MTok | Context | Strength |
|---|---|---|---|---|---|
Eleven Multilingual v2 High-fidelity multilingual TTS Speech | elevenlabs-tts | — | — | TTS | High-fidelity multilingual TTS |
Pricing is indicative — vendors change rates without notice. Routing + fallbacks auto-adjust against each provider's live pricing via LiteLLM. See API reference for the authoritative model list your key can hit right now.
It routes the calls. You set the limits.
The cap holds.
Your organization gets a monthly spend allowance sized to its plan. At the cap, requests stop with a clear error instead of running up an invoice.
One cap covers every key.
The allowance sits on your organization, not on a key, so a new key or a busy agent can’t spend around it.
Your plan picks the model list.
Each plan reaches a set list of models, so a free account can’t call a flagship model by accident.
Keys rotate in one click.
Rotate a key from the dashboard. The provider credentials behind it never leave the gateway.
Only your organization sees its spend.
Keys and spend logs live in their own database schema, and every read checks you belong to the organization.
- Nothing bills past the cap.
- Every key shares one allowance.
- Only your organization sees its spend.
Fair questions
The questions people ask first. Straight answers.
Will this replace my team?
No. It takes away the juggling of five vendor dashboards. Your engineers still choose the model for each job; they just reach every model with one key, and finance gets one line instead of five invoices.
How is this related to Smooth Operator?
It’s the same gateway Smooth Operator speaks through. Your agents’ turns and your own API calls share one key, one cap, and one spend page, and Smooth Operator can tell you which model and which conversation used the most.
Do I have to change my code?
Change the base URL. The gateway speaks the OpenAI API, so the OpenAI SDK, LangChain, and anything else that talks to it keep working.
What happens when we hit the cap?
Requests stop with a 429 and a message that says the budget is spent, until the month rolls over or your plan changes. Nothing bills past it.
Which models can I use?
OpenAI, Anthropic, and Google, plus Groq and direct connections to DeepSeek, Qwen, Kimi, GLM, and MiniMax. The catalog below lists every model with its price and what your plan reaches.
Why not just call the providers directly?
You can, and keep a key, a cap, and an invoice per vendor. Here it’s one of each, next to the rest of your business on Smoo. ChatGPT helps you think. Smooth Operator does the work, inside your business.
Retire four of those five keys. Keep one, with a cap that holds.
Point your SDK at llm.smoo.ai/v1. Usage lands on your Smoo bill, under your plan’s monthly cap.