127.5 105 ±6 $0.30 cache read $0.03 $2.50 1M 46.2 50 ±9 145.6 127.2 111 ±6 $0.14 cache read $0.035 $0.58 262K 52.3 — Not measured — Not measured 127.1 101 ±7 $0.14 cache read $0.028 $0.28 1M 59.4 — Not measured — Not measured 127.0 104 ±5 $0.60 cache read $0.11 $2.20 200K 41.0 61 ±9 143.3 127.0 108 ±6 $0.40 cache read $0.08 $2.00 262K 59.9 55 ±9 — Not measured Inkling FP4 🇺🇸 thinkingmachines/inkling thinkingmachines/inkling Inference by Thinking Machines 126.9 102 ±6 $1.87 cache read $0.374 $4.68 66K 27.9 58 ±9 148.4 126.9 105 ±9 $0.75 cache read $0.20 $3.50 256K 43.6 62 ±9 151.2 Inference by Fireworks AI 126.8 107 ±9 $0.30 cache read $0.06 $1.20 512K 50.9 71 ±8 147.8 125.6 101 ±7 $0.26 $2.60 262K 42.4 56 ±9 — Not measured 125.3 97 ±10 $0.25 cache read $0.025 $1.50 1M 55.5 48 ±9 145 125.2 98 ±13 $3.00 >128k $6.00 cache read $0.75 $15.00 >128k $30.00 cache write $3.00 256K 32.7 — Not measured 146.9 125.2 96 ±10 $1.20 cache read $0.24 $6.00 256K 43.3 41 ±9 142.4 125.0 101 ±7 $0.56 cache read $0.056 $1.68 164K 63.8 — Not measured 146.3 Gemma 4 26B A4B 🇺🇸 deepinfra/google/gemma-4-26B-A4B-it deepinfra/google/gemma-4-26B-A4B-it Inference by DeepInfra 124.9 102 ±8 $0.07 $0.34 262K 56.4 50 ±9 141.6 Inference by Microsoft Azure 124.6 98 ±14 $0.25 cache read $0.025 $2.00 200K 55.1 40 ±9 146 124.4 — Not enough data yet $0.25 cache read $0.05 $2.00 256K — Not enough data yet — Not measured — Not measured 124.3 100 ±8 stale $0.23 cache read $0.20 $2.30 262K 40.6 — Not measured — Not measured 124.2 101 ±8 $1.25 >200k $2.50 cache read $0.20 $2.50 >200k $5.00 cache write $1.25 1M 39.3 48 ±9 — Not measured Nemotron 3 Ultra 🇺🇸 doubleword/nvidia-nemotron-3-ultra doubleword/nvidia-nemotron-3-ultra Inference by Doubleword 124.1 101 ±12 $0.50 $2.50 262K 33.1 53 ±9 147 o4 Mini 🇺🇸 azure/o4-mini@eastus2 · 🇺🇸 eastus2 azure/o4-mini@eastus2 · 🇺🇸 eastus2 Inference by Microsoft Azure 124.0 101 ±14 $1.10 cache read $0.275 $4.40 200K 56.2 53 ±16 145.2 123.1 96 ±22 $0.60 $2.20 205K 46.8 48 ±9 — Not measured Grok 4.1 Fast 🇺🇸 Newer version xai/grok-4-1-fast-non-reasoning · non-reasoning xai/grok-4-1-fast-non-reasoning · non-reasoning Inference by xAI 122.9 95 ±14 $0.20 >128k $0.40 cache read $0.05 $0.50 >128k $1.00 cache write $0.20 2M 59.7 56 ±9 — Not measured MiniMax M2.5 🇨🇳 bedrock/minimax-m2.5@eu-central-1 · 🇩🇪 eu-central-1 bedrock/minimax-m2.5@eu-central-1 · 🇩🇪 eu-central-1 Inference by Amazon Web Services 122.8 100 ±6 $0.36 $1.44 128K 48.7 64 ±9 — Not measured Inference by Amazon Web Services 122.6 95 ±6 $1.00 cache read $0.10 $5.00 cache write $1.25 200K 35.0 59 ±10 142.7 122.2 91 ±16 $0.20 >128k $0.40 cache read $0.05 $0.50 >128k $1.00 cache write $0.20 2M 65.9 — Not measured 144.3