LLM tokens per second — live multi-provider benchmark

Real inference speed, measured continuously. Every row is a live model from Ollama or OpenCode Go — sorted by tokens per second, benchmarked every ~10 minutes on Ollama Pro and every ~30 minutes on OpenCode Go for its standard-tier models.

Independent & vendor-neutral — no paid benchmark rankings. TokenDyno is not affiliated with any provider it benchmarks and receives no payment for ranking or favorable coverage. Read the independence policy →

● live — last benchmark 42s ago
Trend
Gemma4 31B Ollama Pro 325.3 313.2 332ms 100% 4.4 4m ago
GPT-OSS 120B Ollama Pro 309.1 247.1 346ms 100% 3.8 8m ago
Muse Spark Open Ollama Pro 302.1 86.1 932ms 100% 24.9 8m ago
Step 5 Preview OpenCode Go 276.3 525.7 2.6s 100% n=4 28.5 1h ago
MiMo V2.6 Flash Ollama Pro 237.6 106.4 1.4s 100% 26.7 1m ago
MiMo V2.6 Pro Ollama Pro 223.2 181.4 406ms 100% 29.9 1m ago
Inkling 2 Small Ollama Pro 221.6 138.1 470ms 99% 30.6 3m ago
GLM 5.3 Flash Ollama Pro 205.9 229.5 320ms 100% 27.8 8m ago
GLM 5.3 Ollama Pro 194.9 192.3 654ms 100% 31.7 8m ago
DeepSeek V4.1 Flash 1017 Ollama Pro 189.6 140.8 721ms 100% 32.6 8m ago
GLM 5.2 Ollama Pro 188.6 210.5 412ms 99% — 6m ago
DeepSeek V4.1 Flash Ollama Pro 187.5 137.9 376ms 100% — 8m ago
Claude Haiku 5 5 OpenCode Go 179.7 173.1 569ms 100% n=4 36.5 1h ago
Beam (Free) Free model tokens within the paid OpenCode Go plan, according to its pricing at build time. Promotions may end. OpenCode Go 162.4 153.3 455ms 100% n=4 — 1h ago
Beam Ollama Pro 156.1 159.5 500ms 100% 17.7 6m ago
DeepSeek V4 Pro 0813 Ollama Pro 156.1 183.8 405ms 100% 23.1 3m ago
GPT-OSS 20B Ollama Pro 152.4 179.3 406ms 100% — 3m ago
MiniMax M3 Ollama Pro 138.7 135.4 714ms 100% 18.9 8m ago
Kimi K2.7 Code Ollama Pro 138.4 184.4 645ms 100% — 1m ago
MiniMax M3 OpenCode Go 129.0 170.7 678ms 100% n=4 18.9 1h ago
Nemotron 3 Super Ollama Pro 121.4 60.5 4.8s 100% — 7m ago
Kimi K2.6 Ollama Pro 118.0 152.0 489ms 100% — 1m ago
GPT 6 Luna OpenCode Go 111.8 123.4 678ms 100% n=4 27.9 37m ago
GPT 5.6 Luna OpenCode Go 103.3 113.9 853ms 100% n=4 23.8 1h ago
Iceman Contributor (Free) Free model tokens within the paid OpenCode Go plan, according to its pricing at build time. Promotions may end. OpenCode Go 102.4 94.5 3.8s 100% n=4 — 1h ago
Hy3 OpenCode Go 100.1 96.3 2.2s 100% n=4 — 1h ago
GLM 5.3 OpenCode Go 98.6 104.0 656ms 100% n=4 31.7 1h ago
MiniMax M3.1 Flash Ollama Pro 93.7 84.4 1.1s 100% 29.1 7m ago
Muse Spark 1.2 Contributor OpenCode Go 91.1 94.4 2.6s 100% n=4 — 2h ago
Muse Spark 1.3 Contributor OpenCode Go 89.6 122.8 2.3s 75% n=4 36.6 1h ago
GLM 5.2 OpenCode Go 89.2 100.5 557ms 100% n=4 — 1h ago
Mistral Large 4 Ollama Pro 82.0 51.1 1.5s 100% 26.1 4m ago
Hy4 OpenCode Go 81.3 83.6 1.3s 100% n=4 34.6 1h ago
DeepSeek V4.1 Flash OpenCode Go 78.6 77.7 1.5s 100% n=4 — 2h ago
DeepSeek V4 Pro OpenCode Go 77.1 83.4 1.0s 100% n=4 23.1 27m ago
Kimi K3.1 Pro Ollama Pro 72.9 45.2 603ms 99% 46.5 5m ago
Kimi K3 Ollama Pro 72.6 108.1 671ms 100% n=6 — 1h ago
GLM 5.3 Flash OpenCode Go 69.1 82.8 616ms 100% n=4 27.8 1h ago
Kimi K2.7 Code OpenCode Go 63.8 80.7 1.3s 100% n=4 — 1h ago
MiMo V2.6 Flash OpenCode Go 62.5 59.5 3.9s 100% n=4 26.7 1h ago
Gemini 4 Argon OpenCode Go 61.2 56.0 2.5s 100% n=4 50.4 1h ago
Grok 4.7 OpenCode Go 60.0 60.7 1.7s 100% n=4 35.5 1h ago
Kimi K3 OpenCode Go 59.2 66.6 1.7s 100% n=4 — 56m ago
Qwen3.7 Plus OpenCode Go 57.9 50.5 1.6s 100% n=4 — 4h ago
Qwen3.8 Flash OpenCode Go 55.2 61.8 5.1s 100% n=4 — 1h ago
Longcat 2.5 (Free) Free model tokens within the paid OpenCode Go plan, according to its pricing at build time. Promotions may end. OpenCode Go 54.5 54.7 1.4s 100% n=4 — 1h ago
Longcat 2.0 OpenCode Go 50.8 53.0 1.6s 100% n=4 — 1h ago
Qwen3.8 Max OpenCode Go 45.0 45.5 869ms 100% n=4 33.7 57m ago
MiMo V2.6 Pro OpenCode Go 43.8 50.0 862ms 100% n=4 29.9 59m ago
Nemotron 3 Ultra Ollama Pro 38.4 37.8 627ms 95% 11.8 42s ago
Hy4 Preview OpenCode Go 26.8 30.1 1.7s 100% n=4 — 1h ago
Kimi K3.1 Pro OpenCode Go 24.8 22.5 1.7s 100% n=4 46.5 1h ago

Intelligence Index scores from Artificial Analysis.

Ollama Pro is sampled about every 10 minutes. OpenCode Go is sampled about every 30 minutes for its standard-tier models, while its most expensive models — and any Go model not yet cleared for the faster tier — are held at about every 6 hours; sampling may pause or slow below those configured cadences while the shared Go plan budget is conserved.

1 model unavailable or stale

Support the live benchmarks

Frequently asked questions

What is the best LLM tokens per second leaderboard?

There is no single "best" leaderboard because different tools measure different things. TokenDyno is a live tokens-per-second tracker: it re-benchmarks Ollama Pro and OpenCode Go models on a fixed schedule — about every 10 minutes on Ollama Pro and about every 30 minutes on OpenCode Go for its standard-tier models — using the same prompt and measurement method across providers, so speed numbers are directly comparable. For a capability-only score, see LiveBench; for a combined capability-speed-price view, see Artificial Analysis. See the /vs page for a factual side-by-side.

What is TokenDyno's benchmark methodology?

TokenDyno sends a fixed streaming chat-completion request (max_tokens capped at 300) to each provider and measures generation throughput using a hybrid method: inter-token timing when a response streams smoothly (≥8 chunks, mean gap >50ms), or a wall-clock fallback for burst/chunked delivery. Time-to-first-token, error taxonomy, and reliability are also recorded. Full detail is on the /methodology page.

How often is TokenDyno's data refreshed?

The worker benchmarks continuously using a round-robin priority queue. Cadence follows how each provider bills: our Ollama Pro key is a legacy flat-rate subscription, so it is sampled about every 10 minutes. OpenCode Go is sampled about every 30 minutes for its standard-tier models, while its most expensive models — and any Go model not yet cleared for the faster tier — are held at about every 6 hours. Go sampling may pause or slow below those configured cadences while the shared Go plan budget is being conserved (the meter reflects account-level usage, which may include other usage beyond this benchmark). The leaderboard also polls the API roughly every 60 seconds client-side to keep cells current without a full page reload.