Models
Compare scores and official API list prices, not marketplace quotes. Open a model name for the full page — scores, where to buy, and routing. Agent scores use a heat spark so rank is scannable; extended benches stay hidden until you ask. Use Find providers when you already know which models you want.
Sorted by TB (highest first). Click a column header to change.
| # | Status | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | OpenAIGPT-5.6 Sol | $5 | $30 | 88.8★ | 64.6(v) | ? | 69 | ~1M | Proprietary | ★ TB 2.1 SOTA |
| 2 | xAIGrok 4.6 | $2 | $6 | 88.4AA | ? | ? | ? | 500K | Proprietary | ★ AA TB 88.4 · $6/M |
| 3 | MoonshotKimi K3 | $3.00 | $15.00 | 88.3★ open | ? | 84.2 | 62 | 1M | Kimi K3 License | Largest open model |
| 4 | Z.AIGLM-5.3 | ? | ? | 88.2(v) · TB3 28.3 | ? | ? | 115AA | 1M | Proprietary | ★ CyberGym SOTA · TB3 28.3 · Coding Plan |
| 5 | AnthropicClaude Mythos 5 | $10 | $50 | 88(v) | 77.8(v) | ? | ? | 1M+ | Proprietary (restricted) | Glasswing only |
| 6 | OpenAIGPT-5.6 Terra | $2.50 | $15 | 87.4 | 63.4(v) | ? | 141 | ~1M | Proprietary | ★ Half Sol cost, 2-3pts behind |
| 7 | AlibabaQwen3.8 Max | $2 | $6 | 86.6★ (v) | 67.7(v) | ? | 47AA | 1M | Qwen3.8-Max License | ★ Beats Opus TB · 2.4T weights out |
| 8 | OpenAIGPT-5.6 Luna | $1 | $6 | 84.7 | 62.7(v) | ? | 216 | ~1M | Proprietary | ★ Best $/TB in frontier |
| 9 | AnthropicClaude Fable 5 | $10 | $50 | 84.3(v) | 80(v) | 83.3 | 59 | 1M+ | Proprietary | Classifier gap |
| 10 | xAIGrok 4.5 | $2 | $6 | 83.3 | 64.7(v) | ? | 80 | 500K | Proprietary | ★ Cost-efficient frontier |
| 11 | DeepSeekDeepSeek V4 Flash | $0.14 | $0.28 | 82.7★ official 0731 | 52.6(v) | 69 | 100 | 1M | MIT | ★ TB 2.1 82.7 · cheapest frontier agent |
| 12 | ProprietarySWE-1.7 | FREE | FREE | 81.5(v) | 77.8(m) | ? | ? | — | Proprietary | ★ Free preview in Devin · FC 42.3 |
| 13 | Z.AIGLM-5.2 | $0.76 | $2.42 | 81 | 62.1(v) | 76.8 | 109 | 1M | MIT | Open-weight prior flagship |
| 14 | AnthropicClaude Sonnet 5 | $2 | $10 | 80.4★ | 63.2 | ? | 79 | 1M | Proprietary | ★ Beats Opus on TB |
| 15 | MetaMuse Spark 1.1 | $1.25 | $4.25 | 80 | 61.5 | 88.1★★ | 114 | 1M | Proprietary | ★ MCP Atlas leader · Meta API |
| 16 | AnthropicClaude Opus 4.8 | $5 | $25 | 79~ | 69.2(v) | 77.8 | 56 | 1M | Proprietary | ★ SWE-Pro leader |
| 17 | GoogleGemini 3.6 Flash | $1.50 | $7.50 | 78 | 58.7(v) | ? | ? | 1M | Proprietary | Token-efficient workhorse |
| 18 | GoogleGemini 3.5 Flash | $1.50 | $9 | 76.2 | 55.1(v) | 83.6★ | 156 | 1M+ | Proprietary | Best MCP flash |
| 19 | AlibabaQwen3.8-27B | ? | ? | 73(v) | 61.7(v) | ? | ? | 262K | Apache 2.0 | ★ Open 27B · beats Plus on TB |
| 20 | TencentHy3 | $0.13 | $0.53 | 71.7 | 57.9 | 79.1 | 169 | 256K | Apache 2.0 | Open budget agent |
| 21 | ByteDanceSeed 2.1 Pro | $0.83 | $4.14 | 71(v) | ? | ? | ? | 256K | Proprietary | Vendor TB 71 · China API |
| 22 | AlibabaQwen3.7 Max | $1.25 | $3.75 | 69.7 | 60.6 | 76.4 | 197 | 1M | Proprietary | ★ 35h proven |
| 23 | CursorComposer 2.5 | $0.50 | $2.50 | 69.5 | 79.8 | ? | ? | 200K | Proprietary | ★ 1/10th frontier cost, Cursor-only |
| 24 | DeepSeekDeepSeek V4 Pro | $0.44 | $0.87 | 67.9 | 76.2(a) | 73.6 | 64 | 1M | MIT | DeepSWE 8% |
| 25 | ByteDanceSeed 2.1 Turbo | $0.41 | $2.07 | 67.6(v) | ? | ? | ? | 256K | Proprietary | Faster Seed 2.1 |
| 26 | MoonshotKimi K2.7 Code | $0.95 | $4 | 67.4 | 62~ | 76 | 48 | 256K | Modified MIT License | MCP-Mark 81.1 ★ |
| 27 | MiniMaxMiniMax M3 | $0.30 | $1.20 | 66(v) | 59(v) | 74.2 | 111 | 1M | MIT | ★ Open TB leader |
| 28 | XiaomiMiMo v2.5 Pro | $0.17 | $0.34 | 64 | ? | ? | 51 | 1M | MIT | Untested |
| 29 | AlibabaQwen3.7-Plus | $0.32 | $1.28 | 61 | ? | 73.2 | 52 | 1M | Proprietary | Vision Plus · 1M ctx |
| 30 | StepFunStep 3.7 Flash | $0.20 | $1.15 | 59.5 | 56.3 | ? | 387 | 256K | Apache 2.0 | Free tier + multimodal |
| 31 | NVIDIANemotron 3 Ultra | $0.50 | $2.20 | 56.4 | 71.9(v) | ? | 197 | 1M | OpenMDW License v1.1 | 300+ tok/s |
| 32 | GoogleGemini 3.1 Pro | $2 | $12 | 54.2✗ | 46.1(s) | 78.2 | 134 | 2M | Proprietary | 2M context, weak agent |
| 33 | GoogleGemini 3.5 Flash-Lite | $0.30 | $2.50 | 54 | 54.2 | ? | 350 | 1M | Proprietary | Fastest 3.5-series · 350 tok/s |
| 34 | xAIGrok Build 0.1 | $1 | $2 | 52.1 | 70.8(v) | ? | 60 | 256K | Proprietary | Fast coder + Composer |
| 35 | StepFunStep 3.5 Flash | $0.10 | $0.30 | 51(v) | 74.4(v) | ? | 196 | 256K | Apache 2.0 | ★ Cheapest coding |
| 36 | AnthropicClaude Haiku 4.5 | $1 | $5 | 44.2 | 39.5(s) | 40.2 | 100 | 200K | Proprietary | Fast, unverified agent |
| 37 | CustomRing 2.6 1T | $0.30 | $2.50 | 43.1 | ? | ? | 126 | 262K | Custom | Free on OR |
| 38 | AlibabaQwen3-Coder-Next | ~$0.15 | ~$0.60 | 38.2 | 70.6 | ? | 106 | 256K | Apache 2.0 | 3B active only |
| 39 | CohereCohere North Mini Code | FREE | FREE | 35.6(v) | ? | ? | 99 | 256K | Apache 2.0 | Free Apache · 3B active |
| 40 | NVIDIANemotron 3 Super | $0.09 | $0.45 | 31 | 60.47(v) | ? | 146 | 1M | NVIDIA Open Model License Agreement | 12B active MoE |
| 41 | CohereCommand A+ | $2.50 | $10 | 22.8✗ | ? | 57.06 | 162 | 128K | Apache 2.0 | Apache 2.0, weak agent |
| 42 | MistralMistral Large 3 | $0.50 | $1.50 | 12 | 76.2 | 70.4 | 46 | 256K | Apache 2.0 | TB 12% — avoid agents |
| 43 | MetaLlama 4 Maverick | $0.17-0.63 | $0.60-1.79 | 7.9 | 70.4(m) | ? | 118 | 1M | Llama 4 Community License Agreement | tau-bench 93.7 |
| 44 | DeepSeekDeepSeek V4 Pro Max | $1.60 | $3.20 | ? | 55.4 | ? | 64 | 1M | MIT | BenchLM #13 |
| 45 | DeepSeekDeepSeek V4.5 | $0.50 | $1.10 | ? | 62.1 | ? | ? | 256K | MIT | ★ Open SWE-Pro value |
| 46 | BaiduERNIE 5.1 | $0.59 | $2.65 | ? | ? | ? | ? | 128K | Proprietary | Chinese only |
| 47 | ProprietaryAgnes 2.0 Flash | $0.10 | $0.20 | ? | ? | ? | ? | 256K | Proprietary | Unverified |
| 48 | ProprietaryFugu Ultra | sub | sub | ? | 73.7(v) | ? | ? | pool | Proprietary | Multi-agent orchestrator |
| 49 | GoogleDiffusionGemma 26B-A4B | — | — | ? | ? | ? | ? | 256K | Apache 2.0 | Open diffusion decode |
| 50 | GoogleGemma 4 12B | $0.10 | $0.30 | ? | ? | ? | 56 | 256K | Apache 2.0 | Open 12B · $0.30/M |
| 51 | AnthropicClaude Opus 5 | $5.00 | $25.00 | ? | 79.2(v) | 85.8(v) | 56.7AA | 1M | Proprietary | Fable-class at half cost |
| 52 | GoogleGemini 3.5 Flash Cyber | ? | ? | ? | ? | ? | ? | ? | Proprietary | Restricted security pilot |
Pricing Tiers
Budget (<$0.30/M out)
| Agnes 2.0 Flash | $0.20 | Cheapest reasoning |
| DS V4 Flash | $0.28 | TB 2.1 82.7 frontier |
Value ($0.30-1.15/M out)
| Step 3.5F | $0.30 | Cheapest coding/$ |
| Gemma 4 12B | $0.30 | Open 12B |
| MiMo v2.5 Pro | $0.34 | |
| Nemotron 3S | $0.45 | |
| Hy3 | $0.53 | Apache 2.0, TB 2.1 71.7 |
| DS V4 Pro | $0.87 | Single-pass |
| DS V4.5 | $1.10 | Open SWE-Pro 62.1 |
| Step 3.7F | $1.15 | Free tier |
Mid ($1.20-4.40/M out)
| MiniMax M3 | $1.20 | 1M ctx, open TB leader |
| Qwen3.7-Plus | $1.28 | Vision Plus, TB 61 |
| Mistral L3 | $1.50 | |
| Grok Build 0.1 | $2 | Fast coder |
| Seed 2.1 Turbo | $2.07 | Faster Seed 2.1 |
| Nemotron 3U | $2.20 | |
| GLM-5.2 | $2.42 | Open-weight prior flagship |
| Ring 2.6 1T | $2.50 | 1T MoE |
| Composer 2.5 | $2.50 | |
| Gemini 3.5 Flash-Lite | $2.50 | |
| ERNIE 5.1 | $2.65 | Chinese only |
| DS V4 Pro Max | $3.20 | BenchLM #13 |
| Qwen3.7 Max | $3.75 | 35h proven |
| Kimi K2.7 | $4 | MCP-Mark 81.1 |
| Seed 2.1 Pro | $4.14 | Vendor TB 71 |
| Muse Spark 1.1 | $4.25 | MCP Atlas 88.1 leader |
Premium ($5+/M out)
| Claude Haiku 4.5 | $5 | Fast, unverified |
| Luna | $6 | TB 84.7, best $/TB |
| Grok 4.5 | $6 | TB 83.3, cost-efficient |
| Grok 4.6 | $6 | TB 88.4 AA |
| Qwen3.8 Max | $6 | TB 86.6 (v), beats Opus |
| Gemini 3.6 Flash | $7.50 | |
| Gemini 3.5 Flash | $9 | MCP #2 (after Spark) |
| Sonnet 5 | $10 | Beats Opus TB |
| Command A+ | $10 | Apache 2.0, weak TB |
| Gemini 3.1P | $12 | 2M ctx |
| Terra | $15 | TB 87.4, half Sol cost |
| Kimi K3 | $15.00 | |
| Claude Opus 4.8 | $25 | Marathon 26.0 |
| Claude Opus 5 | $25.00 | |
| GPT-5.6 Sol | $30 | TB 88.8 SOTA |
| Claude Fable 5 | $50 | Restored, classifier gap |
| Claude Mythos 5 | $50 |